lang
stringclasses
8 values
year
stringclasses
27 values
title_en
stringlengths
4
253
title_cs
stringlengths
0
251
abstract_en
stringlengths
49
5.12k
abstract_cs
stringlengths
33
3.92k
authors
sequencelengths
1
577
s2_url
stringlengths
79
79
title_fr
stringclasses
3 values
abstract_fr
stringclasses
3 values
title_ru
stringclasses
13 values
abstract_ru
stringclasses
11 values
title_sk
stringclasses
2 values
abstract_sk
stringclasses
2 values
title_de
stringclasses
4 values
abstract_de
stringclasses
4 values
title_dsb
stringclasses
2 values
abstract_dsb
stringclasses
2 values
title_lt
stringclasses
1 value
abstract_lt
stringclasses
1 value
en
2022
Neural Conversational AI
Neuronová konverzační umělá inteligence
A brief introduction to dialogue systems with a specific focus on neural models.
Stručný úvod do dialogových systémů se zvláštním zaměřením na neuronové modely.
[ "Ondřej Dušek" ]
https://www.semanticscholar.org/paper/964c22f1d14c1199130a13db70d85c4d834958d5/
null
null
null
null
null
null
null
null
null
null
null
null
cs
2022
Problems of today's language generators
Problémy dnešních generátorů jazyka
Natural language generation is a task in which a computer uses some input data to produce a comprehensible and fluent message in, for example, English or Czech. How does it work?
Generování přirozeného jazyka je úloha, při níž počítač na základě nějakých vstupních dat vytvoří srozumitelné a plynulé sdělení např. v angličtině nebo češtině. Jak funguje?
[ "Ondřej Dušek" ]
https://www.semanticscholar.org/paper/1563d422f68db44c79f188242de2c94cec607024/
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
Combining Static and Contextualised Multilingual Embeddings
Kombinace statický a kontextuálních embedingů
Static and contextual multilingual embeddings have complementary strengths. Static embeddings, while less expressive than contextual language models, can be more straightforwardly aligned across multiple languages. We combine the strengths of static and contextual models to improve multilingual representations. We extract static embeddings for 40 languages from XLM-R, validate those embeddings with cross-lingual word retrieval, and then align them using VecMap. This results in high-quality, highly multilingual static embeddings. Then we apply a novel continued pre-training approach to XLM-R, leveraging the high quality alignment of our static embeddings to better align the representation space of XLM-R. We show positive results for multiple complex semantic tasks. We release the static embeddings and the continued pre-training code. Unlike most previous work, our continued pre-training approach does not require parallel text.
Statické a kontextové vícejazyčné embeddingy mají komplementární přednosti. Statické embeddingy, i když jsou méně expresivní než kontextové jazykové modely, se dají lépe párovat mezi jazyky. V tomto článku kombinujeme přednosti statických a kontextových, čímž docílíme vyšší kvality vícejazyčných kontextových embedingů. Z modelu XLM-R extrahujeme statické embeddingy pro 40 jazyků, validujeme jejich kvalitu pomocí indukce dvojjazyčných slovníků a pak je zarovnáváme pomocí nástroje VecMap. Výsledkem jsou vysoce kvalitní, vysoce vícejazyčné statické embeddingy. Poté aplikujeme nový přístup pokračujícího pre-training modelu XLM-R, kde využíváme tyto statické embeddingy pro lepší zarovnání reprezentačního prostoru XLM-R. Náš postup dosazuje pozitivních výsledků pro sémanticky náročných úloh. Statické embeddingy a kód pokračujícího pre-training jsou veřejně dostupné. Na rozdíl od většiny předchozí práce náš přístup pokračujícího pre-training nevyžaduje paralelní text.
[ "Katharina Hämmerl", "Jindřich Libovický", "Alexander Fraser" ]
https://www.semanticscholar.org/paper/a763a850b2cd1ee7bc1d51638be0d710406eef34/
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
Computational Literary Studies Infrastructure (CLSINFRA): a H2020 Research Infrastructure Project that aids to connect researchers, data, and methods
Computational Literary Studies Infrastructure (CLSINFRA): Výzkumná infrastruktura H2020, která pomáhá propojit výzkumníky, data a metody
The aim of this poster is to provide an overview of the principal objectives of the newly started H2020 Computational Literary Studies (CLS) project- https://www.clsinfra.io. CLS is a infrastructure project works to develop and bring together resources of high-quality data, tools and knowledge to aid new approaches to studying literature in the digital age. Conducting computational literary studies has a number of challenges and opportunities from multilingual and bringing together distributing information. At present, the landscape of literary data is diverse and fragmented. Even though many resources are currently available in digital libraries, archives, repositories, websites or catalogues, a lack of standardisation hinders how they are constructed, accessed and the extent to which they are reusable (Ciotti 2014). CLS project aims to federate these resources, with the tools needed to interrogate them, and with a widened base of users, in the spirit of the FAIR and CARE principles (Wilkinson et al. 2016). The resulting improvements will benefit researchers by bridging gaps between greater- and lesser- resourced communities in computational literary studies and beyond, ultimately offering opportunities to create new research and insight into our shared and varied European cultural heritage. Rather than building entirely new resources for literary studies, the project is committed to exploiting and connecting the already-existing efforts and initiatives, in order to acknowledge and utilize the immense human labour that has already been undertaken. Therefore, the project builds on recently- compiled high-quality literary corpora, such as DraCor and ELTeC (Fischer et al. 2019, Burnard et al. 2021, Schöch et al. in press), integrates existing tools for text analysis, e.g. TXM, stylo, multilingual NLP pipelines (Heiden 2010, Eder et al. 2016), and takes advantage of deep integration with two other infrastructural projects, namely the CLARIN and DARIAH ERICs. Consequently, the project aims at building a coherent ecosystem to foster the technical and intellectual findability and accessibility of relevant data. The ecosystem consists of (1) resources, i.e. text collections for drama, poetry and prose in several languages, (2) tools, (3) methodological and theoretical considerations, (4) a network of CLS scholars based at different European institutions, (5) a system of short-term research stays for both early career researchers and seasoned scholars, (6) a repository for training materials, as well as (7) an efficient dissemination strategy. This is achieved through a collaboration between participating institutions: Institute of Polish Language at the Polish Academy of Sciences, Poland; University of Potsdam, Germany; Austrian Academy of Sciences, Austria; National University of Distance Education, Spain; École Normale Supérieure de Lyon, France; Humboldt University of Berlin, German; Charles University, Czech Republic; Digital Research Infrastructure for the Arts and Humanities, France; Ghent Centre for Digital Humanities, Ghent University, Belgium; Belgrade Centre for Digital Humanities, Serbia; Huygens Institute for the History of the Netherlands (Royal Netherlands Academy of Arts and Sciences), Netherlands; Trier Center for Digital Humanities, Trier University, Germany; Moore Institute, National University of Ireland Galway, Ireland;
Cílem tohoto posteru je poskytnout přehled hlavních cílů nově zahájeného H2020 Projekt Computational Literary Studies (CLS) – https://www.clsinfra.io. CLS je infrastrukturní projekt pracuje na vývoji a sdružování zdrojů vysoce kvalitních dat, nástrojů a znalostí na pomoc novým přístupy ke studiu literatury v digitálním věku. Vedení počítačových literárních studií má a množství výzev a příležitostí z vícejazyčnosti a spojení distribuce informace. V současnosti je krajina literárních dat různorodá a roztříštěná. I když mnozí zdroje jsou v současnosti dostupné v digitálních knihovnách, archivech, úložištích, webových stránkách nebo katalozích, nedostatek standardizace brání tomu, jak jsou konstruovány, zpřístupněny a do jaké míry jsou znovu použitelné (Ciotti 2014). Cílem projektu CLS je spojit tyto zdroje s nástroji potřebnými k jejich dotazování, a s rozšířenou základnou uživatelů, v duchu principů FAIR a CARE (Wilkinson et al. 2016). Výsledná zlepšení budou přínosem pro výzkumníky tím, že překlenou mezery mezi většími a menšími komunity v oblasti počítačových literárních studií i mimo ni, které v konečném důsledku nabízejí příležitosti vytvořit nový výzkum a vhled do našeho sdíleného a rozmanitého evropského kulturního dědictví. Spíše než budování zcela nových zdrojů pro literární vědu se projekt zavázal k využívání a propojování již existujících snah a iniciativ za účelem uznání a využití obrovská lidská práce, která již byla vykonána. Proto projekt navazuje na nedávné sestavil vysoce kvalitní literární korpusy, jako jsou DraCor a ELTeC (Fischer et al. 2019, Burnard et al. 2021, Schöch a kol. v tisku), integruje stávající nástroje pro analýzu textu, např. TXM, stylo, vícejazyčné NLP potrubí (Heiden 2010, Eder et al. 2016) a využívá výhody hluboké integrace se dvěma dalšími infrastrukturní projekty, jmenovitě konsorcia CLARIN a DARIAH ERIC. V důsledku toho si projekt klade za cíl budování koherentního ekosystému pro podporu technické a intelektuální nalezitelnosti a dostupnosti relevantní údaje. Ekosystém se skládá z (1) zdrojů, tedy sbírek textů pro drama, poezii a prózu několik jazyků, (2) nástroje, (3) metodologické a teoretické úvahy, (4) síť CLS vědeckých pracovníků působících v různých evropských institucích, (5) systém krátkodobých výzkumných pobytů pro oba brzy kariérní výzkumníci a zkušení učenci, (6) úložiště školicích materiálů a také (7) an efektivní strategie šíření. Toho je dosaženo prostřednictvím spolupráce mezi zúčastněnými instituce: Ústav polského jazyka při Polské akademii věd, Polsko; univerzita Postupim, Německo; Rakouská akademie věd, Rakousko; Národní univerzita distančního vzdělávání, Španělsko; École Normale Supérieure de Lyon, Francie; Humboldtova univerzita v Berlíně, německy; Charlesi univerzita, Česká republika; Digitální výzkumná infrastruktura pro umění a humanitní vědy, Francie; Gent Centrum digitálních humanitních věd, Univerzita v Gentu, Belgie; Bělehradské centrum pro digitální humanitní vědy, Srbsko; Huygensův institut pro dějiny Nizozemska (Královská nizozemská akademie umění a věd), Holandsko; Trier Center for Digital Humanities, Trier University, Německo; Mooreův institut, národní Irská univerzita Galway, Irsko;
[ "Julie Birkholz", "Ingo Börner", "Silvie Cinková", "Sally Chambers", "Karina van Dalen-Oskam", "Tess Dejaeghere", "Julia Dudar", "Maciej Eder", "Jennifer Edmond", "Vicky Garnett", "Michal Křen", "Michał Mrugalski", "Ciara Lynn Murphy", "Carolin Odebrecht", "Eliza Papaki", "Marco Raciti", "Lisanne van Rossum", "Christof Schöch", "Srishti Sharma", "Justin Tonra", "Erszsébet Tóth-Czifra", "Peer Trilcke" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
CLS Infra Computational Literary Studies Infrastructure
Infrastruktura CLS pro počítačovou literární vědu
Computational Literary Studies Infrastructure, funded by the Horizon2020 grant scheme, is a four-year, pan-European project that aims to unify the diverse landscape of computational text analysis, in terms of available texts, tools, methods, practices and so forth, within its growing international user community. The project started out in February 2021, meaning that it has been underway for just over a year. In our poster we discuss the various deliverables and activities that have come out of the CLS INFRA project in its first quarter to give an idea of its impact in practice.
Computational Literary Studies Infrastructure, financovaný z grantového schématu Horizont 2020, je čtyřletý celoevropský projekt, jehož cílem je sjednotit rozmanité prostředí počítačové analýzy textů, pokud jde o dostupné texty, nástroje, metody, postupy atd. její rostoucí mezinárodní komunita uživatelů. Projekt byl zahájen v únoru 2021, což znamená, že probíhá něco málo přes rok. V našem plakátu diskutujeme o různých výstupech a aktivitách, které vzešly z projektu CLS INFRA v jeho prvním čtvrtletí, abychom si udělali představu o jeho dopadu v praxi.
[ "Julie Birkholz", "Ingo Börner", "Sally Chambers", "Vera Charvat", "Silvie Cinková", "Karina van Dalen-Oskam", "Tess Dejaeghere", "Julia Dudar", "Matej Ďurčo", "Jennifer Edmond", "Evgeniia Fileva", "Frank Fischer", "Serge Heiden", "Michal Křen", "Bartłomiej Kunda", "Michał Mrugalski", "Ciara Lynn Murphy", "Carolin Odebrecht", "Marco Raciti", "Salvador Ros", "Christof Schöch", "Artjoms Šeļa", "Toma Tasovac", "Justin Tonra", "Erszsébet Tóth-Czifra", "Peer Trilcke", "Maciej Eder", "Lisanne van Rossum" ]
https://www.semanticscholar.org/paper/9a20a70bb033461098d7ad6a116a1f8f1ecef8c0/
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
FINDINGS OF THE IWSLT 2022 EVALUATION CAMPAIGN
Výsledky evaluační kampaně IWSLT 2022
The evaluation campaign of the 19th International Conference on Spoken Language Translation featured eight shared tasks: (i) Simultaneous speech translation, (ii) Offline speech translation, (iii) Speech to speech translation, (iv) Low-resource speech translation, (v) Multilingual speech translation, (vi) Dialect speech translation, (vii) Formality control for speech translation, (viii) Isometric speech translation. A total of 27 teams participated in at least one of the shared tasks. This paper details, for each shared task, the purpose of the task, the data that were released, the evaluation metrics that were applied, the submissions that were received and the results that were achieved.
Hodnoticí kampaň 19. Mezinárodní konference o překladu z mluveného jazyka (IWSLT 2021) letos představila osm sdílených úkolů: (i) Simultánní překlad mluvené řeči, (ii) Offline překlad mluvené řeči, (iii) Překlad řeči na řeč, (iv) Překlad řeči s malým množstvím zdrojů, (v) Vícejazyčný překlad řeči, (vi) Překlad řeči v dialektu, (vii) Kontrola formality pro překlad řeči, (viii) Izometrický překlad řeči. Celkem se zúčastnilo 27 týmů alespoň na jedné z úloh. Tento dokument u každého sdíleného úkolu podrobně popisuje účel úkolu, zveřejněná data, použité metriky hodnocení, obdržené podání a dosažené výsledky.
[ "Antonios Anastasopoulos", "Loïc Barrault", "Luisa Bentivogli", "Marcely Zanon Boito", "Ondřej Bojar", "Roldano Cattoni", "Anna Currey", "Georgiana Dinu", "Kevin Duh", "Maha Elbayad", "Clara Emmanuel", "Yannick Estève", "Marcello Federico", "Christian Federmann", "Souhir Gahbiche", "Hongyu Gong", "Roman Grundkiewicz", "Barry Haddow", "Benjamin Hsu", "Dávid Javorský", "Věra Kloudová", "Surafel Melaku Lakew", "Xutai Ma", "Prashant Mathur", "Paul McNamee", "Kenton Murray", "Maria Nadejde", "Satoshi Nakamura", "Matteo Negri", "Jan Niehues", "Xing Niu", "John Ortega", "Juan Pino", "Elizabeth Salesky", "Yun Tang", "Matthias Sperber", "Sebastian Stuker", "Katsuhito Sudoh", "Marco Turchi", "Yogesh Virkar", "Alex Waibel", "Changhan Wang", "Shinji Watanabe" ]
https://www.semanticscholar.org/paper/d78b3af53e690f094747b72d42dbc84b45f571a9/
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
Hausa Visual Genome: A Dataset for Multi-Modal English to Hausa Machine Translation
Hausa Visual Genome: Dataset pro multimodální překlad mezi angličtinou a hausa
Multi-modal Machine Translation (MMT) enables the use of visual information to enhance the quality of translations. The visual information can serve as a valuable piece of context information to decrease the ambiguity of input sentences. Despite the increasing popularity of such a technique, good and sizeable datasets are scarce, limiting the full extent of their potential. Hausa, a Chadic language, is a member of the Afro-Asiatic language family. It is estimated that about 100 to 150 million people speak the language, with more than 80 million indigenous speakers. This is more than any of the other Chadic languages. Despite a large number of speakers, the Hausa language is considered low-resource in natural language processing (NLP). This is due to the absence of sufficient resources to implement most NLP tasks. While some datasets exist, they are either scarce, machine-generated, or in the religious domain. Therefore, there is a need to create training and evaluation data for implementing machine learning tasks and bridging the research gap in the language. This work presents the Hausa Visual Genome (HaVG), a dataset that contains the description of an image or a section within the image in Hausa and its equivalent in English. To prepare the dataset, we started by translating the English description of the images in the Hindi Visual Genome (HVG) into Hausa automatically. Afterward, the synthetic Hausa data was carefully post-edited considering the respective images. The dataset comprises 32,923 images and their descriptions that are divided into training, development, test, and challenge test set. The Hausa Visual Genome is the first dataset of its kind and can be used for Hausa-English machine translation, multi-modal research, and image description, among various other natural language processing and generation tasks.
Multimodální strojový překlad umožňuje zapojit do překladu vizuální informaci. Článek představuje dataset pro experimenty s multimodálním překladem mezi angličtinou a jazykem hausa.
[ "Idris Abdulmumin", "Satya Ranjan Dash", "Musa Abdullahi Dawud", "Shantipriya Parida", "Shamsddeen Muhammad", "Ibrahim Sa'id Ahmad", " Subhadarshi Panda", "Ondřej Bojar", "Bashir Shehu Galadanci", "Bello Shehu Bello" ]
https://www.semanticscholar.org/paper/69b7585cde35aab63d9bd77a3d7df06e8ca659e9/
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
A Balanced Data Approach for Evaluating Cross-Lingual Transfer: Mapping the Linguistic Blood Bank
Vyvážený datový přístup pro hodnocení translingvistického přenosu: Mapování lingvistické krevní banky
We show that the choice of pretraining languages affects downstream cross-lingual transfer for BERT-based models. We inspect zero-shot performance in balanced data conditions to mitigate data size confounds, classifying pretraining languages that improve downstream performance as donors, and languages that are improved in zero-shot performance as recipients. We develop a method of quadratic time complexity in the number of languages to estimate these relations, instead of an exponential exhaustive computation of all possible combinations. We find that our method is effective on a diverse set of languages spanning different linguistic features and two downstream tasks. Our findings can inform developers of large-scale multilingual language models in choosing better pretraining configurations.
Ukazujeme, že výběr jazyků pro předškolení ovlivňuje následný přenos mezi jazyky pro modely založené na BERT. Zkoumáme nulový výkon v vyvážených datových podmínkách, abychom zmírnili matoucí velikost dat, klasifikujeme jazyky pro předškolení, které zlepšují následný výkon, jako dárce a jazyky, které se zlepšují v nulovém výkonu, jako příjemce. Vyvíjíme metodu kvadratické časové složitosti v počtu jazyků, abychom tyto vztahy odhadli, místo exponenciálního vyčerpávajícího výpočtu všech možných kombinací. Zjišťujeme, že naše metoda je účinná na různorodou sadu jazyků zahrnující různé jazykové vlastnosti a dva navazující úkoly. Naše zjištění mohou informovat vývojáře rozsáhlých vícejazyčných jazykových modelů při výběru lepších konfigurací pro předškolení.
[ "Dan Malkin", "Tomasz Limisiewicz", "Gabriel Stanovsky" ]
https://www.semanticscholar.org/paper/0c158af75176a98cafa8507b76c67beaf80119a7/
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
Reflexives as Part of Verb Lexemes in the VALLEX Lexicon
Reflexiva tvořící součást lexému a jejich popis ve slovníku VALLEX
Reflexivity represents one of the core research tasks in current linguistics. As the use of reflexives, encoding a variety of meanings, typically brings about changes in verb valency, the description of reflexivity is highly relevant – among others – also for valency oriented studies. In this paper, we address the reflexive in Czech categorized as a derivational morpheme (e.g., zlomit ‘to break something’ --> zlomit se ‘to break; to crack’), with the focus on valency behavior of reflexive verbs as represented in the valency lexicon of Czech verbs VALLEX. In the data component of the lexicon, reflexive verbs, i.e., verbs with reflexive lexemes, are captured in separate lexicon entries, represented by respective verb lemma(s) containing the free reflexive morpheme se or si. In VALLEX, there are 922 lexical entries for reflexive verbs described in 1 545 lexical units represented by 1 525 verb lemmas (this number covers almost one quarter of lexical units and one third of verb lemmas in the lexicon). Reflexive verbs can be divided into two groups: into those without any non-reflexive counterpart (reflexiva tantum, 208 lexical units represented by 177 verb lemmas) and into those for which a non-reflexive base verb can be identified (derived reflexive verbs, 1 337 lexical units represented by 1 348 verb lemmas). Those derived reflexive verbs that are directly related to their non-reflexive base verbs are classified into seven types on the ground of their relation to the non-reflexive counterparts, captured in the data component of the lexicon by the value of the attribute reflexverb. Further, the relation of derived reflexive verbs and their respective non-reflexive counterparts is described by formal rules comprised in the grammar component of the lexicon (19 rules in total), which provide the information on changes in the mapping of semantic participants onto valency complementations.
Adekvátní popis reflexivity patří ke stěžejním výzkumným úkolům současné lingvistiky. Přítomnost reflexiva obvykle přináší změny valence, proto je popis reflexivity vysoce relevantní m.j. též pro studie zaměřené na slovesnou valenci. V této práci se zabýváme reflexivními slovesy v češtině, která lze charakterizovat jako slovesa obsahující slovotvorný morfém (např. zlomit --> zlomit se), se zaměřením na valenční chování těchto sloves a jejich popis ve valenčním lexikonu českých sloves VALLEX. V datové komponentě slovníku jsou reflexivní slovesa zachycena v samostatných slovníkových záznamech, které jsou reprezentovány příslušnými slovesnými lemmaty (obsahujícími volný reflexivní morfém se nebo si). Ve VALLEXu identifikujeme 922 lexikálních záznamů pro reflexivní slovesa, které popisují celkem 1 545 lexikálních jednotek reprezentovaných 1 525 slovesmými lemmaty (toto číslo pokrývá téměř jednu čtvrtinu lexikálních jednotek a jednu třetinu slovesných lemmat ve slovníku). Reflexivní slovesa lze rozdělit do dvou skupin: na slovesa bez nereflexivního základového slovesa (reflexiva tantum, 208 lexikálních jednotek reprezentovaných 177 slovesnými lemmaty) a na slovesa, u nichž lze identifikovat nereflexivní základní sloveso (odvozená reflexivní slovesa, 1 337 lexikálních jednotek reprezentovaných 1 348 slovesnými lemmaty). Odvozená reflexivní slovesa, která jsou přímo odvozena od nereflexivních základových sloves, řadíme do sedmi typů na základě jejich vztahu k nereflexivním základovým slovesům (v datové komponentě je tento vztah zachycen v atributu reflexverb). Vztah odvozených reflexních sloves a jejich příslušných nereflexních protějšků je dále popsán formálními pravidly obsaženými v gramatické složce lexikonu (celkem 19 pravidel), která poskytují informace o změnách v mapování sémantických participantů na odpovídající valenční doplnění.
[ "Václava Kettnerová", "Markéta Lopatková", "Anna Vernerová" ]
https://www.semanticscholar.org/paper/724aa6461a765dd0c5aab904312270e9ba4778df/
null
null
null
null
null
null
null
null
null
null
null
null
cs
2022
Permeation (Czech premiere of a theatre play)
Prostoupení (česká premiéra divadelní hry)
Czech premiere of the second theatre play generated within the THEaiTRE project.
Česká premiéra druhé divadelní hry vygenerované v rámci projektu THEaiTRE.
[ "Josef Doležal", "2.0 THEaiTRobot", "Daniel Hrbek", "Rudolf Rosa", "Tomáš Studeník" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
cs
2022
Discussion after the double performance of AI: When a Robot Writes a Play and Permeation
Diskuze po dvojpředstavení AI: Když robot píše hru a Prostoupení
What does the robot think about, how does it feel and can it write a play? Snippets from the life of artificial intelligence in its own words. Futuristic Little Prince of the postpandemic era. Every reprise will be followed by a debate with experts in artificial intelligence and theater makers. The show is not only a testament to the current capabilities of computer technology, but also an engaging vision of the future world inspired by the sci-fi classics.
O čem robot přemýšlí, co cítí a dokáže napsat divadelní hru? Střípky ze života umělé inteligence jejími vlastními slovy. Futuristický Malý princ z doby postpandemické. Inscenace, po jejíž každé repríze následuje debata s odborníky na umělou inteligenci i divadelními tvůrci, je nejen svědectvím o současných schopnostech počítačových technologií, ale i poutavou vizí budoucího světa inspirující se v klasikách žánru sci-fi.
[ "Josef Doležal", "Daniel Hrbek", "Rudolf Rosa" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
Multimodality for NLP-Centered Applications: Resources, Advances and Frontiers
Multimodalita pro aplikace zpracovní přirozeného jazyka: Zdroje, pokroky, hranice
With the development of multimodal systems and natural language generation techniques, the resurgence of multimodal datasets has attracted significant research interests, which aims to provide new information to enrich the representation of textual data. However, there remains a lack of a comprehensive survey for this task. To this end, we take the first step and present a thorough review of this research field. This paper provides an overview of a publicly available dataset with different modalities according to the applications. Furthermore, we discuss the new frontier and give our thoughts. We hope this survey of multimodal datasets can provide the community with quick access and a general picture of the multimodal dataset for specific Natural Language Processing (NLP) applications and motivates future researches. In this context, we release the collection of all multimodal datasets easily accessible here: https://github.com/drmuskangarg/Multimodal-datasets
Přehledový článek představuje první krok k souhrnné prezentaci veřejně dostupných datasetů pro multimodální experimenty.
[ "Muskan Garg", "Seema Wazarkar", "Muskaan Singh", "Ondřej Bojar" ]
https://www.semanticscholar.org/paper/5acf711385b23b963b91dca11b3ff5deffa46505/
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
Combining Noisy Semantic Signals with Orthographic Cues: Cognate Induction for the Indic Dialect Continuum
Kombinace sémantických a pravopisných rysů pro detekci kognátů v indickém dialektovém kontinuu
We present a novel method for unsupervised cognate/borrowing identification from monolingual corpora designed for low and extremely low resource scenarios, based on combining noisy semantic signals from joint bilingual spaces with orthographic cues modelling sound change. We apply our method to the North Indian dialect continuum, containing several dozens of dialects and languages spoken by more than 100 million people. Many of these languages are zero-resource and therefore natural language processing for them is non-existent. We first collect monolingual data for 26 Indic languages, 16 of which were previously zero-resource, and perform exploratory character, lexical and subword cross-lingual alignment experiments for the first time at this scale on this dialect continuum. We create bilingual evaluation lexicons against Hindi for 20 of the languages. We then apply our cognate identification method on the data, and show that our method outperforms both traditional orthography baselines as well as EM-style learnt edit distance matrices. To the best of our knowledge, this is the first work to combine traditional orthographic cues with noisy bilingual embeddings to tackle unsupervised cognate detection in a (truly) low-resource setup, showing that even noisy bilingual embeddings can act as good guides for this task. We release our multilingual dialect corpus, called HinDialect, as well as our scripts for evaluation data collection and cognate induction.
Představujeme novou neřízenou metodu pro hledání kognátů v monolinguálních datech při scénářích s malými a extrémně malými daty. Metoda kombinuje zašumělý sémantický signál ze sdruženého bilinguálního prostoru s pravopisnými rysy modelujícími zvukové změny. Metodu aplikujeme na severoindické dialektové kontinuum, které pokrývá několik desítek jazyků a dialektů. Pro mnoho z těchto jazyků neexistovaly žádné datové zdroje, nástroje pro zpracování přirozeného jazyka pro tyto jazyky tedy v podstatě také neexistovaly. Nejprve jsme shromáždili monolinguální korpusy pro 26 indických jazyků, z nichž pro 16 dosud neexistovaly datové zdroje, a provedli jsme na nich řadu pilotních experimentů. Pro 20 jazyků jsme vytvořili překladové slovníky do hindštiny. Následně jsme aplikovali naši metodu pro hledání kognátů a doložili jsme, že její úspěšnost překonává klasické systémy založené jenom na pravopisných rysech i metody založené na řetězcové podobnosti natrénované způsobem EM. Jsme přesvědčeni, že naše práce jako první kombinuje využití pravopisných rysů se zašumělými bilinguálními embeddingy a umožňuje hledání kognátů v situacích se skutečně kritickým nedostatkem dat. Naši vícejazyčnou datovou sadu jsme zveřejnili pod názvem HinDialect, zdrojové kódy jsme zveřejnili také.
[ "Niyati Bafna", "Josef Genabith", "Cristina España-Bonet", "Zdeněk Žabokrtský" ]
https://www.semanticscholar.org/paper/4c634317136f6ce4fd18e5aa176ddf91ea5be0f1/
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
Constrained Decoding for Technical Term Retention in English-­Hindi MT
Omezené dekódování pro zachování odborných termínů v anglicko-hindském MT
Technical terms may require special handling when the target audience is bilingual, depending on the cultural and educational norms of the society in question. In particular, certain translation scenarios may require “term retention” i.e. preserving of the source language technical terms in the target language output to produce a fluent and comprehensible code­switched sentence. We show that a standard transformer­based machine translation model can be adapted easily to perform this task with little or no damage to the general quality of its output. We present an English­to­Hindi model that is trained to obey a “retain” signal, i.e. it can perform the required code­switching on a list of terms, possibly unseen, provided at runtime. We perform automatic evaluation using BLEU as well as F1 metrics on the list of retained terms; we also collect manual judgments on the quality of the output sentences.
Technické termíny mohou vyžadovat zvláštní zacházení, pokud je cílová skupina dvojjazyčná, v závislosti na kulturních a vzdělávacích normách dané společnosti. Zejména některé překladové situace mohou vyžadovat „zachování termínů“, tj. ponechání technických termínů zdrojového jazyka v cílovém jazyce výstupu, aby vznikla plynulá a srozumitelná věta. Ukazujeme, že standardní model strojového překladu založený na Transformeru lze snadno přizpůsobit k dosažení tohoto cíle, aniž by přitom trpěla kvalita jeho výstupu obecně. Představujeme anglicko-hindský model, který je natrénovat k uposlechnutí signálu „zachování“, tj. za běhu zajistí, že vybrané termíny nebudou přeloženy. Navrženou metodu vyhodnocujeme automatickými metrikami (BLEU pro překlad obecně, F1 pro zachování termínů), i ručně (celková kvalita výstupních vět).
[ "Niyati Bafna", "Martin Vastl", "Ondřej Bojar" ]
https://www.semanticscholar.org/paper/9e69660130d126b14118510dfb4a8a5c876ebc08/
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
Subword-based Cross-lingual Transfer of Embeddings from Hindi to Marathi and Nepali
Přenos slovních embeddingů z hindštiny do maratštiny a nepálštiny na základě podslovných jednotek
Word embeddings are growing to be a crucial resource in the field of NLP for any language. This work introduces a novel technique for static subword embeddings transfer for Indic languages from a relatively higher resource language to a genealogically related low resource language. We primarily work with Hindi-Marathi, simulating a low-resource scenario for Marathi, and confirm observed trends on Nepali. We demonstrate the consistent benefits of unsupervised morphemic segmentation on both source and target sides over the treatment performed by fastText. Our best-performing approach uses an EM-style approach to learning bilingual subword embeddings; we also show, for the first time, that a trivial “copy- and-paste” embeddings transfer based on even perfect bilingual lexicons is inadequate in capturing language-specific relationships. We find that our approach substantially outperforms the fastText baselines for both Marathi and Nepali on the Word Similarity task as well as WordNet-Based Synonymy Tests; on the former task, its performance for Marathi is close to that of pretrained fastText embeddings that use three orders of magnitude more Marathi data.
Slovní embeddingy jsou ve zpracování přirozeného jazyka čím dál podstatnější komponentou. Tento článek představuje novou metodu pro přenos statických podslovných embeddings z jazyka s relativním dostatkem zdrojů do jazyka s nedostatkem zdrojů. Primárně pracujeme s jazykovým párem hindština-márátština, přičemž pro márátštinu je nedostatek zdrojů pouze simulovaný a výsledky dále potvrzujme na nepálštině. Náš přístup výrazně překonává baseline fastText pro oba jazyky na úlohách podobnost slov a testování synonymie. Na první úloze je úspěšnost na máráštině dokonce srovnatelná s úspěšností standardní metody při využití o tři řády většího množství dat. We primarily work with Hindi-Marathi, simulating a low-resource scenario for Marathi, and confirm observed trends on Nepali. We demonstrate the consistent benefits of unsupervised morphemic segmentation on both source and target sides over the treatment performed by fastText. Our best-performing approach uses an EM-style approach to learning bilingual subword embeddings; we also show, for the first time, that a trivial “copy-and-paste” embeddings transfer based on even perfect bilingual lexicons is inadequate in cap- turing language-specific relationships. We find that our approach substantially outperforms the fastText baselines for both Marathi and Nepali on the Word Similarity task as well as WordNet-Based Synonymy Tests; on the former task, its performance for Marathi is close to that of pretrained fastText embeddings that use three orders of magnitude more Marathi data.
[ "Niyati Bafna", "Zdeněk Žabokrtský" ]
https://www.semanticscholar.org/paper/4328a6ad84517306e4c8d423c94690ade44c196b/
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
Overview of the ELE Project
Přehled projektu ELE
This paper presents the ongoing European Language Equality (ELE) project, an 18-month action funded by the European Commission. The primary goal of the ELE project is to prepare the ELE programme, in the form of a strategic research, innovation and implementation agenda and roadmap for achieving full digital language equality in Europe by 2030.
Tento dokument představuje probíhající projekt Evropské jazykové rovnosti (ELE), což je osmnáctiměsíční akce financovaná Evropskou komisí. Primárním cílem projektu ELE je připravit program ELE v podobě strategického programu výzkumu, inovací a realizace a plánu pro dosažení úplné digitální jazykové rovnosti v Evropě do roku 2030.
[ "Itziar Aldabe", "Jane Dunne", "Aritz Farwell", "Owen Gallagher", "Federico Gaspari", "Maria Giagkou", "Jan Hajič", "Jens Peter Kückens", "Teresa Lynn", "Georg Rehm", "German Rigau", "Katrin Marheinecke", "Stelios Piperidis", "Natalia Resende", "Tereza Vojtěchová", "Andy Way" ]
https://www.semanticscholar.org/paper/0aef113a6c2facc4fde2915deea3dcc2f123a373/
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
Migrant Stories
Příběhy migrantů
Migrant Stories is a corpus of 1017 short biographic narratives of migrants supplemented with meta information about countries of origin/destination, the migrant gender, GDP per capita of the respective countries etc. The corpus has been compiled as a teaching material for data analysis.
Příběhy migrantů je korpus 1017 krátkých biografických příběhů migrantů doplněných metainformacemi o zemích původu/cílových zemích, pohlaví migrantů, HDP na obyvatele příslušných zemí apod. Korpus byl vytvořen jako výukový materiál pro analýzu dat.
[ "Martin Hájek", "Jiří Mírovský", "Barbora Hladká" ]
https://www.semanticscholar.org/paper/6bcad32ee51ced9273555b475287586bf7a82f67/
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
GEMv2: Multilingual NLG Benchmarking in a Single Line of Code
GEMv2: Vícejazyčný NLG Benchmarking v jediném řádku kódu
Evaluation in machine learning is usually informed by past choices, for example which datasets or metrics to use. This standardization enables the comparison on equal footing using leaderboards, but the evaluation choices become sub-optimal as better alternatives arise. This problem is especially pertinent in natural language generation which requires ever-improving suites of datasets, metrics, and human evaluation to make definitive claims. To make following best model evaluation practices easier, we introduce GEMv2. The new version of the Generation, Evaluation, and Metrics Benchmark introduces a modular infrastructure for dataset, model, and metric developers to benefit from each others work. GEMv2 supports 40 documented datasets in 51 languages. Models for all datasets can be evaluated online and our interactive data card creation and rendering tools make it easier to add new datasets to the living benchmark.
Vyhodnocování ve strojovém učení je obvykle založeno na minulých rozhodnutích, například na tom, které datové sady nebo metriky použít. Tato standardizace umožňuje rovnocenné porovnávání pomocí žebříčků, ale volby hodnocení se stávají suboptimálními, protože se objevují lepší alternativy. Tento problém se týká zejména generování přirozeného jazyka, které vyžaduje stále se zdokonalující sady datových sad, metrik a lidské hodnocení, aby bylo možné učinit definitivní závěry. Abychom usnadnili sledování nejlepších postupů vyhodnocování modelů, představujeme GEMv2. Nová verze benchmarku pro generování, vyhodnocování a metriky zavádí modulární infrastrukturu pro vývojáře datových sad, modelů a metrik, aby mohli vzájemně využívat své práce. GEMv2 podporuje 40 zdokumentovaných datových sad v 51 jazycích. Modely pro všechny datové sady lze vyhodnocovat online a naše interaktivní nástroje pro tvorbu a vykreslování datových karet usnadňují přidávání nových datových sad do živého benchmarku.
[ "Sebastian Gehrmann", "Abhik Bhattacharjee", "Abinaya Mahendiran", "Alex Wang", "Alexandros Papangelis", "Aman Madaan", "Angelina McMillan-Major", "Anna Shvets", "Ashish Upadhay", "Bingsheng Yao", "Bryan Wilie", "Chandra Bhagavatula", "Chaobin You", "Craig Thomson", "Cristina Garbacea", "Dakuo Wang", "Daniel Deutsch", "Deyi Xiong", "Di Jin", "Dimitra Gkatzia", "Dragomir R. Radev", "Elizabeth Clark", "Esin Durmus", "Faisal Ladhak", "Filip Ginter", "Genta Indra Winata", "Hendrik Strobelt", "Hiroaki Hayashi", "Jekaterina Novikova", "Jenna Kanerva", "Jenny Chim", "Jiawei Zhou", "Jordan Clive", "Joshua Maynez", "Joao Sedoc", "Juraj Juraska", "Kaustubh D. Dhole", "Khyati Rangavi Chandu", "Laura Perez-Beltrachini", "Leonardo F. R. Ribeiro", "Lewis Tunstall", "Li Zhang", "Mahima Pushkarna", "Mathias Creutz", "Michael White", "Mihir Kale", "Moussa Kamal Eddine", "Nico Daheim", "Nishant Subramani", "Ondřej Dušek", "Paul Pu Liang", "Pawan Sasanka Ammanamanchi", "Qi Zhu", "Ratish Puduppully", "Reno Kriz", "Rifat Shahriyar", "Ronald Cardenas", "Saad Mahamood", "Salomey Osei", "Samuel Cahyawijaya", "Sanja Štajner", "Sebastien Montella", "Shailza Jolly", "Simon Mille", "Tahmid Hasan", "Tianhao Shen", "Tosin Adewumi", "Vikas Raunak", "Vipul Raheja", "Vitaly Nikolaev", "Vivian Tsai", "Yacine Jernite", "Ying Xu", "Yisi Sang", "Yixin Liu", "Yufang Hou" ]
https://www.semanticscholar.org/paper/9ed047bb4ef6ddc296d473bf4f3b55488aeba350/
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
Two Reproductions of a Human-Assessed Comparative Evaluation of a Semantic Error Detection System
Dvě reprodukce lidmi hodnoceného srovnávacího hodnocení systému detekce sémantických chyb
In this paper, we present the results of two reproduction studies for the human evaluation originally reported by Dušek and Kasner (2020) in which the authors comparatively evaluated outputs produced by a semantic error detection system for data-to-text generation against reference outputs. In the first reproduction, the original evaluators repeat the evaluation, in a test of the repeatability of the original evaluation. In the second study, two new evaluators carry out the evaluation task, in a test of the reproducibility of the original evaluation under otherwise identical conditions. We describe our approach to reproduction, and present and analyse results, finding different degrees of reproducibility depending on result type, data and labelling task. Our resources are available and open-sourced.
V tomto článku prezentujeme výsledky dvou reprodukčních studií lidského hodnocení, které původně reportovali Dušek a Kasner (2020) a ve kterém autoři srovnávali výstupy vytvořené systémem pro detekci sémantických chyb při generování textu z dat s referenčními výstupy. V první reprodukční studii původní hodnotitelé opakují hodnocení v rámci testu opakovatelnosti původního hodnocení. Ve druhé studii dva noví hodnotitelé provádějí úlohu hodnocení v rámci testu reprodukovatelnosti původního hodnocení za jinak stejných podmínek. Popisujeme náš přístup k reprodukci a prezentujeme a analyzujeme výsledky, přičemž zjišťujeme různé stupně reprodukovatelnosti v závislosti na typu výsledku, údajích a úloze označování. Naše zdroje jsou k dispozici a mají otevřený zdrojový kód.
[ "Rudali Huidrom", "Ondřej Dušek", "Zdeněk Kasner", "Thiago Castro Ferreira", "Anya Belz" ]
https://www.semanticscholar.org/paper/317804972a8976a6606c8a245e3b4f6bbb4bdae6/
null
null
null
null
null
null
null
null
null
null
null
null
cs
2022
Rudolf Rosa: We have proved that AI is almost advanced enough to write a theater play
Rudolf Rosa: Dokázali jsme, že doba, kdy AI zvládne napsat divadelní hru, je velmi blízko
He is the scientist who introduced artificial intelligence to the stage. Nevertheless, he admits that the machine may not create art because it has no inner need to express anything. Read about how Rudolf Rosa, a researcher at the Institute of Formal and Applied Linguistics, CUNI's MFF, sees the future of AI in art.
Umělou inteligenci dostal na divadelní prkna. Přesto přiznává, že stroj možná nemůže tvořit umění, protože nemá vnitřní potřebu cosi vyjádřit. Přečtěte si, kde vidí budoucnost AI v umění Rudolf Rosa, výzkumný pracovník z Ústavu formální a aplikované lingvistiky MFF UK.
[ "Petr Pohůdka", "Rudolf Rosa" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
Rudolf Rosa: We have proved that AI is almost advanced enough to write a theater play
Rudolf Rosa: Dokázali jsme, že doba, kdy AI zvládne napsat divadelní hru, je velmi blízko
He is the scientist who introduced artificial intelligence to the stage. Nevertheless, he admits that the machine may not create art because it has no inner need to express anything. Read about how Rudolf Rosa, a researcher at the Institute of Formal and Applied Linguistics, CUNI's MFF, sees the future of AI in art.
Umělou inteligenci dostal na divadelní prkna. Přesto přiznává, že stroj možná nemůže tvořit umění, protože nemá vnitřní potřebu cosi vyjádřit. Přečtěte si, kde vidí budoucnost AI v umění Rudolf Rosa, výzkumný pracovník z Ústavu formální a aplikované lingvistiky MFF UK.
[ "Petr Pohůdka", "Rudolf Rosa" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
Annotation guidelines for Spanish verbal synonyms in the SynSemClass lexicon
Manual pro anotaci španělských slovních synonym ve slovníku SynSemClass
These guidelines are created as part of the research project Language Understanding: from Syntax to Discourse (LUSyD) and describe the goals, resources and procedure for the addition of a fourth language, Spanish, to the multilingual synonym verbal lexicon SynSemClass. The lexicon currently comprises three languages: Czech, English and German, thus allowing a comparison of syntactic-semantic cross-linguistic properties. The lexicon also provides empirical data relevant for ongoing NLP research.
Tyto pokyny jsou vytvořeny v rámci výzkumného projektu Porozumění jazyku: od syntaxe k diskurzu (LUSyD) a popisují cíle, zdroje a postup pro doplnění čtvrtého jazyka, španělštiny, do SynSemClass. Na stránkách lexikon v současné době obsahuje tři jazyky: češtinu, angličtinu a němčinu, což umožňuje porovnání syntakticko-sémantických vlastností napříč jazyky. Lexikon rovněž poskytuje empirická data relevantní pro probíhající výzkum NLP.
[ "Cristina Fernández Alcaina", "Eva Fučíková", "Zdeňka Urešová" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
Conversion/zero-affixation: theories, data
Konverze: teorie, data
After a good two decades when only a stronghold of very few morphologists defended or even considered the possibility of derivational zeroes (cf. Don 2005), zero-affixation has recently taken the stage of workshops at international conferences again, both past (Iordăchioaia & Melloni, SLE 2020) and to come (Vasile, forthcoming). The discussion aroused by these and other events as well as by recent publications on the issue (e.g. Dahl & Fábregas 2018) has taken the form of revisions, both theory-oriented, e.g. in Gaeta (2013) or ten Hacken (2019), and data-driven, e.g. in Melloni & Iordăchioaia (2020) or Ševčíková (2021), and brings to the fore past controversies and renewed insights on the process and on its most adequate interpretation.
Po dobrých dvou desetiletích, kdy možnost derivačních nul obhajovalo nebo o ní dokonce uvažovalo jen několik málo morfologů (srov. Don 2005), se nulová afixace v poslední době opět dostává na scénu workshopů na mezinárodních konferencích, a to jak minulých (Iordăchioaia & Melloni, SLE 2020), tak budoucích (Vasile, forthcoming). Diskuse vyvolaná těmito a dalšími akcemi i nedávnými publikacemi na toto téma (např. Dahl & Fábregas 2018) má podobu revizí, a to jak teoretických, např. v Gaeta (2013) nebo ten Hacken (2019), tak datových, např. v Melloni & Iordăchioaia (2020) nebo Ševčíková (2021), a přináší do popředí minulé kontroverze i nové pohledy na tento proces a na jeho nejadekvátnější interpretaci.
[ "Cristina Fernández Alcaina", "Salvador Valera" ]
https://www.semanticscholar.org/paper/29608965311139498f8c3b0ffd7b1a97a03b251c/
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
How “Tools” Produce “Data”: Searching in a Large Digital Corpus of Audiovisual Holocaust Testimonies
Jak "nástroje" produkují "data": Vyhledávání ve velkém digitálním korpusu audiovizuálních svědectví
The field of Jewish Studies is facing many new challenges as a result of ongoing digitization. This chapter focuses on digital oral histories of the Hol-ocaust. Following the digital revolution in oral history, many institutions now provide access to multiple collections at once. One of the new challenges is thus related to the simultaneous availability of several archives, as well as various search engines which apply different methods to browse their content. The aim of this chapter is to identify and describe participants’ practices for working with a large corpus of audiovisual Holocaust testimonies, especially in terms of locating relevant results within the collection by using three different search systems. We have conducted an empirical study in an experimental setting designed to emulate work with various search engines. Three pairs of novice users solved ten tasks over video-conferencing software, utilizing three different search “tools” (USC Shoah Foundation’s Visual History Archive, Amalach, and Pixla). Our main find-ings consist of formulating a fundamental structure and elements of participants’ collaborative work, composed of three complementary actions: testing, sharing, and implementing. Furthermore, users obtained the search results by two main approaches: aggregation and query refinement. Interestingly, they did not upgrade the searching skills progressively, but rather used the current “best knowledge” for all the tasks and search engines at once. The participants’ emergent competence was continuously developed on the basis of collaborative work with the search engines and the results obtained so far through their work on the previous tasks.
Židovská studia čelí v důsledku probíhající digitalizace mnoha novým výzvám. Tato kapitola se zaměřuje na digitální ústní dějiny holokaustu. Po digitální revoluci v ústní historii poskytuje nyní mnoho institucí přístup k více sbírkám najednou. Jedna z nových výzev tak souvisí se současnou dostupností několika archivů a také různých vyhledávačů, které používají různé metody pro prohlížení jejich obsahu. Cílem této kapitoly je identifikovat a popsat postupy účastníků při práci s velkým korpusem audiovizuálních svědectví o holokaustu, zejména pokud jde o vyhledávání relevantních výsledků v rámci sbírky pomocí tří různých vyhledávacích systémů. Provedli jsme empirickou studii v experimentálním prostředí určeném k emulaci práce s různými vyhledávači. Tři páry začínajících uživatelů vyřešily deset úkolů přes videokonferenční software, přičemž využily tři různé vyhledávací „nástroje“ (archiv vizuální historie nadace USC Shoah Foundation, Amalach a Pixla). Naše hlavní nálezy spočívají ve formulaci základní struktury a prvků spolupráce účastníků, složené ze tří vzájemně se doplňujících akcí: testování, sdílení a implementace. Uživatelé dále získali výsledky vyhledávání pomocí dvou hlavních přístupů: agregace a zušlechťování dotazů. Zajímavé je, že nezvyšovali dovednosti vyhledávání progresivně, ale spíše využívali současné „nejlepší znalosti“ pro všechny úkoly a vyhledávače najednou. Vznikající kompetence účastníků byla průběžně rozvíjena na základě spolupráce s vyhledávači a výsledků, které dosud získali díky své práci na předchozích úkolech.
[ "Jakub Mlynář", "Jiří Kocián", "Karin Hofmeisterová" ]
https://www.semanticscholar.org/paper/d50d538c7a9be53ccf2ed502abe305e10c5565c7/
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
Prague Discourse Treebank 3.0
Pražský diskurzní korpus 3.0
An updated version of discourse annotation in the Prague Dependency Treebank data, containing numerous corrections of errors and annotation inconsistencies. The data are published in their native format (PDT-like), as well as newly in the Penn Discourse Treebank 3.0 column format and the PDTB3 sense taxonomy.
Aktualizovaná verze diskurzní anotace v datech Pražského závislostního korpusu obsahující opravu řady chyb a anotačních nekonzistencí. Data jsou publikována ve svém nativním formátu (stejný jako u PDT) a nově rovněž ve sloupcovém formátu Penn Discourse Treebanku 3.0 a s diskurzními typy taxonomie PDTB3.
[ "Pavlína Synková", "Magdaléna Rysová", "Jiří Mírovský", "Lucie Poláková", "Veronika Sheller", "Jana Zdeňková", "Šárka Zikánová", "Eva Hajičová" ]
https://www.semanticscholar.org/paper/9224f1a99f10cee8eeef6bbd20aa58e370959e1d/
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
Can we use contextual word embeddings to create word-formational networks?
Lze využít kontextové slovní embeddingy pro tvorbu slovotvorných sítí?
In recent years, research of word embeddings has been focused on contextual embeddings, which, in comparison with static embeddings, naturally provide strong cross-lingual properties and improve empirical results in many downstream tasks. However, most of their uses (named entity recognition, syntactic parsing, translation etc.) benefit from the representation of the context itself, while inherently lexical tasks seem to not benefit as much. In this presentations, I will compare contextual embeddings obtained from XLM-RoBERTa to static word2vec and FastText embeddings and discuss several methods of extracting static representations from the contextual ones.
V posledních letech se výzkum word embeddings zaměřuje na kontextové embeddingy, které v porovnání se statickými embeddingy umožňují snadné rozšíření na více jazyků a dosažení lepších výsledků na úlohách, ve kterých jsou používány. Většina těchto úloh (rozpoznávání pojmenovaných entit, syntaktický větný rozbor, překlad atp.) však těží z reprezentace samotného kontextu, zatímco na lexikálních úlohách kontextové embeddingy přinášejí pouze malé zlepšení. Ve své prezentaci srovnám kontextové embeddingy z XLM-RoBERTa se statickými embeddingy natrénovanými pomocí word2vec a FastText a ukáži několik metod pro extrakci statických reprezentací z kontextových.
[ "Jonáš Vidra" ]
https://www.semanticscholar.org/paper/572720f0a026cfd2d3f5baa663e4c5ddd77520e6/
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
Permeation (premiere of a theatre play)
Prostoupení (premiéra divadelní hry)
Can a robot write a play that reflects the world in the times we live in? Find out at the stage reading of a play created fully by a computer. Follow the story of the married couple Ivan and Nina, whose lives become deeply impacted by incoming war. Permeation is the second play generated by artificial intelligence—a mix of love tragedy, war drama, and espionage thriller.
Může robot napsat hru, která odráží svět v době, ve které žijeme? Zjistěte si to na jevišti při čtení hry vytvořené plně počítačem. Sledujte příběh manželského páru Ivana a Niny, jejichž životy jsou hluboce ovlivněny přicházející válkou. Pronikání je druhá hra vytvořená umělou inteligencí – směs milostné tragédie, válečného dramatu a špionážního thrilleru.
[ "Erwin Maas", "2.0 THEaiTRobot", "Josef Doležal", "Klára Vosecká", "David Košťák", "Rudolf Rosa", "Tomáš Studeník" ]
https://www.semanticscholar.org/paper/bda9d762aeedbcb21f4cc09b80e0e3c2b0f119e0/
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
Discussion after premiere of Permeation
Diskuze po premiéře Prostoupení
Artists, scientists and university students in Prague joined their forces for the second time to explore creative collaboration between a man and artificial intelligence. The stage reading, presented at the 2022 Rehearsal for Truth Theater Festival in world premiere, is outcome of two years of work on a unique project supported by The Technology Agency of the Czech Republic (TACR). At the start of this project there was a question whether AI can write a theater play. Using artificial intelligence during creative process and collaboration of a man and AI poses numerous unsettling questions beyond the field of art and invites for wider social debate.
Umělci, vědci a vysokoškolští studenti v Praze již podruhé spojili své síly, aby prozkoumali tvůrčí spolupráci mezi člověkem a umělou inteligencí. Scénografické čtení, prezentované na divadelním festivalu Rehearsal for Truth v roce 2022 ve světové premiéře, je výsledkem dvouleté práce na unikátním projektu podporovaném Technologickou agenturou České republiky (TAČR). Na začátku tohoto projektu byla otázka, zda AI dokáže napsat divadelní hru. Využití umělé inteligence během tvůrčího procesu a spolupráce člověka a AI vyvolává řadu znepokojivých otázek mimo oblast umění a vybízí k širší společenské debatě.
[ "Erwin Maas", "David Košťák", "Rudolf Rosa" ]
https://www.semanticscholar.org/paper/e5e2f15405efa80951bdfb06a6c81d3a9e55fef0/
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
The Interaction of Personal Data, Intellectual Property and Freedom of Expression in the Context of Language Research
Interakce osobních dat, intelektuálního vlastnictví a svobody slova v kontextu jazykového výzkumu
Language researchers are usually aware of intellectual property and personal data (PD) requirements. The problem, however, arises when these two legal regimes have conflicting requirements. For instance, when copyright law requires the acknowledgement of the author, but personal data law enshrines the data mini-misation principle. It is a practical question for a language researcher whether he should name the author of the text used for, e.g., building a language model, or follow the data minimisation principle not to name the author. The access right that a data subject has introduces similar conflicts. The question is what the scope of the access right is. Does it cover only processed personal data, or does it extend to data derived from PD? The interaction of the freedom of expression with PD protection entails several problems. The question is whether researchers can publish their research results containing personal data. The General Data Protection Regulation establishes a general framework that needs to be implemented by EU member states. We analyse different implementations based on examples from several EU countries.
Jazykoví badatelé si obvykle uvědomují požadavky na duševní vlastnictví a osobní údaje (PD). Problém však nastává, když tyto dva právní režimy mají protichůdné požadavky. Například když autorské právo vyžaduje uznání autora, ale zákon o osobních údajích zakotvuje zásadu minimisace údajů. Pro jazykového badatele je praktickou otázkou, zda by měl jmenovat autora textu použitého např. pro vytvoření jazykového modelu, nebo se řídit zásadou minimalizace údajů, aby autora nejmenoval. Přístupové právo, které má subjekt údajů, přináší podobné konflikty. Otázkou je, jaký je rozsah přístupového práva. Vztahuje se pouze na zpracovávané osobní údaje, nebo se vztahuje i na údaje odvozené od PD? Vzájemné působení svobody projevu s ochranou PD přináší několik problémů. Otázkou je, zda mohou badatelé zveřejňovat výsledky svého výzkumu obsahující osobní údaje. Obecné nařízení o ochraně osobních údajů stanoví obecný rámec, který musí členské státy EU zavést. Analyzujeme různé implementace na základě příkladů z několika zemí EU.
[ "Aleksei Kelli", "Krister Lindén", "Paweł Kamocki", "Kadri Vider", "Penny Labropoulou", "Ramūnas Birštonas", "Vadim Mantrov", "Vanessa Hannesschläger", "Ricardo Del Grata", "Age Värv", "Gaabriel Tavits", "Andres Vutt", "Ester Hoorn", "Jan Hajič", "Arvi Tavast" ]
https://www.semanticscholar.org/paper/954a8da568c3a19ca5c2d7bd6ac3b785cfa8da03/
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
Word Formation Analyzer for Czech: Automatic Parent Retrieval and Classification of Word Formation Processes
Word Formation Analyzer for Czech: Automatické rozpoznávání základových slov a klasifikace slovotvorných procesů
We present a deep-learning tool called Word Formation Analyzer for Czech, which, given an input lexeme, automatically retrieves the lemma or lemmas from which the input lexeme was formed. We call this task parent retrieval. Furthermore, based on the number of words in the output sequence and its comparison to the input, the input word is classified into one of three categories: compound, derivative or unmotivated. We call this task word formation classification. In the task of parent retrieval, Word Formation Analyzer for Czech achieved an accuracy of 71%. In word formation classification, the tool achieved an accuracy of 87%.
Představujeme nástroj vytvořený na bázi hlubokého učení nazvaný Word Formation Analyzer for Czech, který na základě vstupního lexému vrátí lemma nebo lemmata, jež tvoří jeho základ. Této úloze říkáme rozpoznávání základových slov. Na základě počtu slov ve výstupní sekvenci a jejím srovnání se vstupem lze pak vstup klasifikovat do jedné ze tří kategorií: kompozitum, derivát, nebo nemotivované. Této úloze říkáme slovotvorná klasifikace. V rozpoznávání základových slov dosáhl Word Formation Analyzer for Czech accuracy 71%, v klasifikaci slovotvorných procesů pak 87%.
[ "Emil Svoboda", "Magda Ševčíková" ]
https://www.semanticscholar.org/paper/2facf36c796d32447fb775bb6f71199170be0798/
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
Assessing clinical utility of Machine Learning and Artificial Intelligence approaches to analyze speech recordings in Multiple Sclerosis: A Pilot Study
Vyhodnocení klinické užitečnosti přístupů na bázi strojového učení a umělé inteligence při analýze nahrávek pacientů s roztroušenou sklerózou: pilotní studie
Background An early diagnosis together with an accurate disease progression monitoring of multiple sclerosis is an important component of successful disease management. Prior studies have established that multiple sclerosis is correlated with speech discrepancies. Early research using objective acoustic measurements has discovered measurable dysarthria. Method The objective was to determine the potential clinical utility of machine learning and deep learning/AI approaches for the aiding of diagnosis, biomarker extraction and progression monitoring of multiple sclerosis using speech recordings. A corpus of 65 MS-positive and 66 healthy individuals reading the same text aloud was used for targeted acoustic feature extraction utilizing automatic phoneme segmentation. A series of binary classification models was trained, tuned, and evaluated regarding their Accuracy and area-under-the-curve. Results The Random Forest model performed best, achieving an Accuracy of 0.82 on the validation dataset and an area-under-the-curve of 0.76 across 5 k-fold cycles on the training dataset. 5 out of 7 acoustic features were statistically significant. Conclusion Machine learning and artificial intelligence in automatic analyses of voice recordings for aiding multiple sclerosis diagnosis and progression tracking seems promising. Further clinical validation of these methods and their mapping onto multiple sclerosis progression is needed, as well as a validating utility for English-speaking populations.
Pozadí Včasná diagnóza spolu s přesným sledováním vývoje onemocnění roztroušenou sklerózou je důležitou součástí úspěšného zvládání nemoci. Předchozí studie prokázaly, že roztroušená skleróza souvisí s nesrovnalostmi v řeči. Dřívější výzkumy využívající objektivní akustická měření odhalily měřitelnou dysartrii. Metoda Cílem bylo zjistit potenciální klinickou využitelnost technik založených strojového učení a hlubokého učení / umělé inteligence při diagnostice, extrakci biomarkerů a monitorování progrese roztroušené sklerózy pomocí řečových záznamů. Pro cílenou extrakci akustických příznaků s využitím automatické segmentace fonémů byl použit korpus 65 jedinců s RS a 66 zdravých jedinců čtoucích nahlas stejný text. Byla vytrénována, odladěna a vyhodnocena množina binárních klasifikačních modelů co do jejich přesnosti a AUC (Area Under the Curve). Výsledky Nejlepším modelem se ukázal být Random Forest, který dosáhl přesnosti 0,82 na souboru validačních dat a AUC 0,76 v 5 k-krát opakovaných cyklech na souboru trénovacích dat. Pět ze sedmi akustických rysů bylo statisticky významných. Závěr Využití strojové učení a umělé inteligence v automatických analýze hlasových záznamů pro účely diagnostiky roztroušené sklerózy a sledování její progrese se jeví být slibným. Je zapotřebí dalšího klinického ověření těchto metod a jejich mapování na progresi roztroušené sklerózy, stejně jako validace využití též metody pro anglicky mluvící populaci.
[ "Emil Svoboda", "Tomáš Bořil", "Jan Rusz", "Tereza Tykalová", "Dana Horáková", "Charles R. G. Guttmann", "Krastan B. Blagoev", "Hiroto Hatabu", "Vladimir I. Valtchinov" ]
https://www.semanticscholar.org/paper/a23a8de9a16d62f9fd783db90d3fc2a9040eefff/
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
AI Technologies for Machine Supervision and Help in a Rehabilitation Scenario
Technologie AI pro strojový dohled a pomoc v rehabilitačním scénáři
We consider, evaluate, and develop methods for home rehabilitation scenarios. We show the required modules for this scenario. Due to the large number of modules, the framework falls into the category of Composite AI. Our work is based on collected videos with high-quality execution and samples of typical errors. They are augmented by sample dialogues about the exercise to be executed and the assumed errors. We study and discuss body pose estimation technology, dialogue systems of different kinds and the emerging constraints of verbal communication. We demonstrate that the optimization of the camera and the body pose allows high-precision recording and requires the following components: (1) optimization needs a 3D representation of the environment, (2) a navigation dialogue to guide the patient to the optimal pose, (3) semantic and instance maps are necessary for verbal instructions about the navigation. We put forth different communication methods, from video-based presentation to chit-chat-like dialogues through rule-based methods. We discuss the methods for different aspects of the challenges that can improve the performance of the individual components. Due to the emerging solutions, we claim that the range of applications will drastically grow in the very near future.
Zvažujeme, hodnotíme a vyvíjíme metody pro scénáře domácí rehabilitace. Uvádíme moduly potřebné pro tento scénář. Vzhledem k velkému počtu modulů spadá rámec do kategorie kompozitní umělé inteligence. Naše práce je založena na shromážděných videích s kvalitním provedením a ukázkami typických chyb. Ty jsou doplněny ukázkovými dialogy o prováděném cvičení a předpokládaných chybách. Studujeme a diskutujeme technologii odhadu polohy těla, dialogové systémy různých druhů a vznikající omezení verbální komunikace. Ukazujeme, že optimalizace kamery a polohy těla umožňuje vysoce přesný záznam a vyžaduje následující komponenty: (1) optimalizace potřebuje 3D reprezentaci prostředí, (2) navigační dialog, který pacienta navede do optimální pozice, (3) pro slovní instrukce o navigaci jsou nezbytné sémantické a instanční mapy. Předkládáme různé metody komunikace, od prezentace založené na videu až po dialogy typu chit-chat prostřednictvím metod založených na pravidlech. Diskutujeme o metodách pro různé aspekty výzev, které mohou zlepšit výkonnost jednotlivých komponent. Díky vznikajícím řešením tvrdíme, že rozsah aplikací se ve velmi blízké budoucnosti drasticky rozšíří.
[ "Gábor Baranyi", "Bruno Carlos Dos Santos Melício", "Zsófia Gaál", "Levente Hajder", "András Simonyi", "Dániel Sindely", "Joul Skaf", "Ondřej Dušek", "Tomáš Nekvinda", "András Lőrincz" ]
https://www.semanticscholar.org/paper/56fabb31790ccdffb37f226c4ea5668b84db73d2/
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
Balancing the Style-Content Trade-Off in Sentiment Transfer Using Polarity-Aware Denoising
Balancování stylu a obsahu při přenosu sentimentu pomocí denoisingu s ohledem na polaritu
Text sentiment transfer aims to flip the sentiment polarity of a sentence (positive to negative or vice versa) while preserving its sentiment-independent content. Although current models show good results at changing the sentiment, content preservation in transferred sentences is insufficient. In this paper, we present a sentiment transfer model based on polarity-aware denoising, which accurately controls the sentiment attributes in generated text, preserving the content to a great extent and helping to balance the style-content trade-off. Our proposed model is structured around two key stages in the sentiment transfer process: better representation learning using a shared encoder and sentiment-controlled generation using separate sentiment-specific decoders. Empirical results show that our methods outperforms state-of-the-art baselines in terms of content preservation while staying competitive in terms of style transfer accuracy and fluency. Source code, data, and all other related details are available on Github (https://github.com/SOURO/polarity-denoising-sentiment-transfer).
Cílem přenosu sentimentu textu je změnit polaritu sentimentu věty (pozitivní na negativní nebo naopak) při zachování jejího obsahu nezávislého na sentimentu. Ačkoli současné modely vykazují dobré výsledky při změně sentimentu, zachování obsahu v přenesených větách je nedostatečné. V tomto článku představujeme model přenosu sentimentu založený na denoisingu zohledňujícím polaritu, který přesně kontroluje atributy sentimentu v generovaném textu, do značné míry zachovává obsah a pomáhá vyvážit kompromis mezi stylem a obsahem. Námi navrhovaný model je strukturován kolem dvou klíčových fází procesu přenosu sentimentu: učení lepší reprezentace pomocí sdíleného enkodéru a generování řízené sentimentem pomocí specifických dekodérů pro každý sentiment. Empirické výsledky ukazují, že naše metody překonávají současné základní metody z hlediska zachování obsahu a zároveň zůstávají konkurenceschopné z hlediska přesnosti a plynulosti přenosu stylu. Zdrojový kód, data a všechny další související podrobnosti jsou k dispozici na Githubu (https://github.com/SOURO/polarity-denoising-sentiment-transfer).
[ "Sourabrata Mukherjee", "Zdeněk Kasner", "Ondřej Dušek" ]
https://www.semanticscholar.org/paper/02eaa81f5d59d2da9159ef00fe733980785bde56/
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
The recent developments in Universal Anaphora Scorer
Nedávný vývoj v Universal Anaphora Scorer
The Universal Anaphora initiative aims to push forward the state of the art in anaphora and anaphora resolution by expanding the aspects of anaphoric interpretation which are or can be reliably annotated in anaphoric corpora, producing unified standards to annotate and encode these annotations, deliver datasets encoded according to these standards, and developing methods for evaluating models carrying out this type of interpretation. Such expansion of the scope of anaphora resolution requires a comparable expansion of the scope of the scorers used to evaluate this work. Last year, we introduce an extended version of the Reference Coreference Scorer (Pradhan et al., 2014) that can be used to evaluate identity anaphora resolution (including singletons, split-antecedents), bridging reference resolution, non-referring expressions and discourse deixis. The scorer has been used in the two recent CODI-CRAC Shared Tasks on Anaphora Resolution in Dialogues. Recently, an extension of the UA scorer that supports also discontinuous markables has been used by Novák et al (2022) in the CRAC 2022 Shared Task on Multilingual Coreference Resolution. In this talk, we will introduce the details about the scorer on scoring the different aspects of anaphora resolutions and how has it been used in recently shared tasks. In addition, we will also discuss the work in progress for the scorers such as mention overlap ratio, anaphor-decomposable score and the adaptation for CRAFT shared task.
Iniciativa Universal Anaphora si klade za cíl posunout vpřed současný stav v oblasti anafory a rozpoznávání anafory rozšířením aspektů anaforické interpretace, které jsou nebo mohou být spolehlivě anotovány v anaforických korpusech, vytvořením jednotných standardů pro anotaci a kódování těchto anotací, dodáním datových souborů kódovaných podle těchto standardů a vytvořením metod pro hodnocení modelů provádějících tento typ interpretace. Takové rozšíření rozsahu rozlišení anafory vyžaduje srovnatelné rozšíření rozsahu hodnotitelů používaných k hodnocení této práce. V loňském roce jsme představili rozšířenou verzi Reference Coreference Scoreru (Pradhan a kol., 2014), která může být použita k hodnocení rozpoznávání anafory identity (včetně singletonů, split-antecedentů), rozpoznávání bridging anafory, neodkazujících výrazů a diskursu deixis. Skórer byl použit ve dvou nedávných soutěžech CODI-CRAC o rozpoznávání anafory v dialozích. Nedávno bylo rozšíření UA scoreru, které podporuje také nesouvislé zmínky, použito Novákem a kol. (2022) v soutěži CRAC 2022 ve vícejazyčném ropoznávání koreference. V tomto příspěvku představíme podrobnosti o scoreru o vyhodnocení různých aspektů rozlišení anafory a o tom, jak byl použit v nedávných soutěžech. Kromě toho budeme také diskutovat o rozpracované práci pro skórery, jako je skóre překrytí zmínek, skóre delitelné po anaforech a adaptace skóre pro soutěž CRAFT.
[ "Juntao Yu", "Michal Novák" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
CorefUD 1.0: Coreference Meets Universal Dependencies
CorefUD 1.0: Setkání koreference a Universal Dependencies
Recent advances in standardization for annotated language resources have led to successful large scale efforts, such as the Universal Dependencies (UD) project for multilingual syntactically annotated data. By comparison, the important task of coreference resolution, which clusters multiple mentions of entities in a text, has yet to be standardized in terms of data formats or annotation guidelines. In this paper we present CorefUD, a multilingual collection of corpora and a standardized format for coreference resolution, compatible with morphosyntactic annotations in the UD framework and including facilities for related tasks such as named entity recognition, which forms a first step in the direction of convergence for coreference resolution across languages.
Nedávný pokrok ve standardizaci anotovaných jazykových zdrojů vedl k úspěšným velkým projektům jako Universal Dependencies (UD), kde se syntakticky anotují data pro mnoho jazyků. Anotace koreference, která spojuje opakované zmínky téže entity v textu a je pro porozumění jazyku velmi důležitá, je zatím standardizačním úsilím relativně nepoznamenaná. V tomto článku prezentujeme CorefUD, mnohojazyčnou sbírku korpusů a standardizovaný formát pro anotaci koreference, kompatibilní s morfosyntaktickou anotací v UD a rozšiřitelný na příbuzné úlohy, jako je rozpoznávání pojmenovaných entit. Jde o první krok směrem ke konvergenci koreferenčních zdrojů napříč jazyky.
[ "Anna Nedoluzhko", "Michal Novák", "Martin Popel", "Zdeněk Žabokrtský", "Amir Zeldes", "Daniel Zeman" ]
https://www.semanticscholar.org/paper/33336cdc37455107ca34636d844ab352e410eb1a/
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
CorefUD 1.0
CorefUD 1.0
CorefUD is a collection of previously existing datasets annotated with coreference, which we converted into a common annotation scheme. In total, CorefUD in its current version 0.2 consists of 17 datasets for 11 languages, and compared to the version 0.2, the file format has been reworked and a number of annotation errors have been fixed.
CorefUD je kolekce existujících korpusů s anotací koreference, které jsme převedli na jednotné anotační schéma. Ve své současné verzi 1.0 CorefUD celkem obsahuje 17 korpusů pro 11 jazyků a v porovnání s verzí 0.2 obsahuje přepracovaný formát souborů a opravy řady chyb.
[ "Anna Nedoluzhko", "Michal Novák", "Martin Popel", "Zdeněk Žabokrtský", "Amir Zeldes", "Daniel Zeman", "Peter Bourgonje", "Silvie Cinková", "Jan Hajič", "Christian Hardmeier", "Pauline Krielke", "Frédéric Landragin", "Ekaterina Lapshinova-Koltunski", "Antònia Martí Antonín", "Marie Mikulová", "Maciej Ogrodniczuk", "Marta Recasens", "Manfred Stede", "Milan Straka", "Svetlana Toldova", "Veronika Vincze", "Voldemaras Žitkus" ]
https://www.semanticscholar.org/paper/33336cdc37455107ca34636d844ab352e410eb1a/
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
ELITR Minuting Corpus: A Novel Dataset for Automatic Minuting from Multi-Party Meetings in English and Czech
ELITR Minuting Corpus: Nový dataset pro automatické zápisky z vícestranných jednání v angličtině a češtině
Taking minutes is an essential component of every meeting, although the goals, style, and procedure of this activity (“minuting” for short) can vary. Minuting is a relatively unstructured writing act and is affected by who takes the minutes and for whom the minutes are intended. With the rise of online meetings, automatic minuting would be an important use-case for the meeting participants and those who might have missed the meeting. However, automatically generating meeting minutes is a challenging problem due to various factors, including the quality of automatic speech recognition (ASR), public availability of meeting data, subjective knowledge of the minuter, etc. In this work, we present the first of its kind dataset on Automatic Minuting. We develop a dataset of English and Czech technical project meetings, consisting of transcripts generated from ASRs, manually corrected, and minuted by several annotators. Our dataset, ELITR Minuting Corpus, consists of 120 English and 59 Czech meetings, covering about 180 hours of meeting content. The corpus is publicly available at http://hdl.handle.net/11234/1-4692 as a set of meeting transcripts and minutes, excluding the recordings for privacy reasons. A unique feature of our dataset is that most meetings are equipped with more than one minute, each created independently. Our corpus thus allows studying differences in what people find important while taking the minutes. We also provide baseline experiments for the community to explore this novel problem further. To the best of our knowledge, ELITR Minuting Corpus is probably the first resource on minuting in English and also in a language other than English (Czech).
Pořizování zápisků je nezbytnou součástí každé schůze, ačkoli cíle, styl a postup této činnosti ("protokolování", jak se říká "zápisům") jsou velmi odlišné). Zápis je relativně nestrukturovaný písemný úkon a je ovlivněn tím, kdo zápis pořizuje a pro koho je zápis pořizován. S nárůstem online schůzek by automatická tvorba zápisů byla důležitou pomocí pro účastníky schůzek i ty, kteří by mohli schůzku zmeškat. Automatické generování zápisů ze schůzí je však náročný problém, který je způsoben různými faktory včetně kvality automatického rozpoznávání řeči (ASR), veřejné dostupnosti dat o schůzkách, subjektivních znalostí zapisovatele atd. V této práci představujeme první soubor dat svého druhu o automatickém protokolování. Vytváříme datovou sadu anglických a českých technických projektových schůzek, které se skládají z přepisů vygenerovaných ASR, ručně opravených a zaprotokolovaných několika anotátory. Náš dataset, ELITR Minuting Corpus, se skládá ze 120 anglických a 59 českých schůzek, které pokrývají přibližně 180 hodin záznamu. Korpus je veřejně dostupný na adrese http://hdl.handle.net/11234/1-4692 jako soubor přepisů schůzí s výjimkou nahrávek z důvodu ochrany osobních údajů. Unikátní vlastností našeho souboru dat je, že většina schůzí je vybavena více než jedním zápisem, přičemž každý z nich byl vytvořen nezávisle. Náš korpus tak umožňuje studovat rozdíly v tom, co lidé považují za důležité. Poskytujeme také základní experimenty pro komunitu, která může tento nový problém dále zkoumat. Podle našeho nejlepšího vědomí je korpus ELITR Minuting pravděpodobně prvním zdrojem zápisků v angličtině a také v jiném jazyce než v angličtině (češtině).
[ "Anna Nedoluzhko", "Muskaan Singh", "Marie Hledíková", "Tirthankar Ghosal", "Ondřej Bojar" ]
https://www.semanticscholar.org/paper/a737b72feba19eba468be96b834b279546179c98/
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
Survey of Low-Resource Machine Translation
Přehled metod strojového překladu s nedostatkem datových zdrojů
We present a survey covering the state of the art in low-resource machine translation. There are currently around 7000 languages spoken in the world and almost all language pairs lack significant resources for training machine translation models. There has been increasing interest in research addressing the challenge of producing useful translation models when very little translated training data is available. We present a high level summary of this topical field and provide an overview of best practices.
Předkládáme přehled o stavu výzkumu v oblasti strojového překladu s nízkými zdroji. V současnosti se na světě mluví asi 7 000 jazyky a téměř všem jazykovým párům chybí významné zdroje pro trénování modelů strojového překladu. Vzrůstá zájem o výzkum zabývající se problémem vytváření užitečných překladatelských modelů, když je k dispozici velmi málo přeložených trénovacích dat. Předkládáme shrnutí této aktuální oblasti na vysoké úrovni a poskytujeme přehled osvědčených postupů.
[ "Barry Haddow", "Rachel Bawden", "Antonio Valerio Miceli Barone", "Jindřich Helcl", "Alexandra Birch" ]
https://www.semanticscholar.org/paper/0857bff865fd772d34566f62e0ff698e7d0717aa/
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
Continuous Rating as Reliable Human Evaluation of Simultaneous Speech Translation
Continuous Rating jako důvěryhodná metoda lidského hodnocení simultánního strojového překladu řeči
Simultaneous speech translation (SST) can be evaluated on simulated online events where human evaluators watch subtitled videos and continuously express their satisfaction by pressing buttons (so called Continuous Rating). Continuous Rating is easy to collect, but little is known about its reliability, or relation to comprehension of foreign language document by SST users. In this paper, we contrast Continuous Rating with factual questionnaires on judges with different levels of source language knowledge. Our results show that Continuous Rating is easy and reliable SST quality assessment if the judges have at least limited knowledge of the source language. Our study indicates users' preferences on subtitle layout and presentation style and, most importantly, provides a significant evidence that users with advanced source language knowledge prefer low latency over fewer re-translations.
Abstrakt je pouze v angličtině. Simultaneous speech translation (SST) can be evaluated on simulated online events where human evaluators watch subtitled videos and continuously express their satisfaction by pressing buttons (so called Continuous Rating). Continuous Rating is easy to collect, but little is known about its reliability, or relation to comprehension of foreign language document by SST users. In this paper, we contrast Continuous Rating with factual questionnaires on judges with different levels of source language knowledge. Our results show that Continuous Rating is easy and reliable SST quality assessment if the judges have at least limited knowledge of the source language. Our study indicates users' preferences on subtitle layout and presentation style and, most importantly, provides a significant evidence that users with advanced source language knowledge prefer low latency over fewer re-translations.
[ "Dávid Javorský", "Dominik Macháček", "Ondřej Bojar" ]
https://www.semanticscholar.org/paper/52a783f6c83900045fdddc037165b60af27a3dad/
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
Automatic minuting: A pipeline method for generating minutes from multi-party meeting proceedings
Automatické zápiskování: Sekvenční metoda generování zápisů ze schůzek více osob
Automatically generating meeting minutes is a challenging yet time-relevant problem in speech and natural language processing. Nowadays, meeting minutes seem more cru- cial than ever due to the manifold rise of on- line meetings. However, automatic minut- ing is not straightforward for various reasons: obtaining transcriptions of sufficient quality, summarizing long dialogue discourse, retain- ing topical relevance and coverage, handling redundancies and small talk, etc. This pa- per presents our investigations on a pipelined approach to automatically generate meeting minutes using a BART model (Bidirectional and Auto-Regressive Transformers) trained on multi-party dialogue summarization datasets. We achieve comparable results with our sim- ple yet intuitive method with respect to pre- vious large and computationally heavy state- of-the-art models. We make our code avail- able at https://github.com/ELITR/ minuting-pipeline.
Článek představuje základní sekvenční metodu automatického generování zápisků z jednání. Transkript se rozdělí na stejně velké kratší úseky, které sumarizuje model zaločený na BARTu.
[ "Kartik Shinde", "Tirthankar Ghosal", "Ondřej Bojar" ]
https://www.semanticscholar.org/paper/98134a7f456b62e4f1fb51a906e94261681668ae/
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
CUNI-Bergamot Submission at WMT22 General Task
Systém UK v General Task na WMT22
We present CUNI-Bergamot submission for WMT22 General translation task. We compete in English → Czech direction. Our submission further explores block backtranslation techniques. In addition to the previous work, we measure performance in terms of COMET score and named entities translation accuracy. We evaluate performance of MBR decoding compared to traditional mixed backtranslation training and we show possible synergy when using both of the techniques simultaneously. The results show that both approaches are effective means of improving translation quality and they yield even better results when combined.
Představujeme systém CUNI-Bergamot pro WMT22 General translation task. Soutěžíme ve směru angličtina → čeština. Naše práce dále zkoumá techniky blokového zpětného překladu. Oproti předchozí práci měříme kvalitu z hlediska skóre COMET a přesnosti překladu pojmenovaných entit. Hodnotíme kvalitu MBR dekódování ve srovnání s tradičním tréninkem smíšeného zpětného překladu a ukazujeme možnou synergii při současném použití obou technik. Výsledky ukazují, že oba přístupy jsou účinným prostředkem ke zlepšení kvality překladu a při jejich kombinaci dosahujeme ještě lepších výsledků.
[ "Josef Jon", "Martin Popel", "Ondřej Bojar" ]
https://www.semanticscholar.org/paper/9fc017e9f6d465d0bf5e6862baf65664e60d7b6a/
null
null
null
null
null
null
null
null
null
null
null
null
cs
2022
Contrastive analysis of frequency peaks of written and spoken lexicon
Kontrastivní analýza frekvenčních špiček psaného a mluveného lexika
A set of partial analyses of semantically and formally-grammatically defined groups of lexemes (adverbs, deictics, numerals, proper nouns, diminutives, female gender counterparts, and univerbized forms) belonging to the frequency peak of the spoken or written Czech language. A dataset of 3000 of the most frequent lemmas in the ORAL v1 and ORTOFON v1 corpora, or SYN2015; the differences between the two sets can be seen as a manifestation of diglossia in Czech at the lexical level.
Soubor dílčích analýz sémanticky a formálně-gramaticky vymezených skupin lexémů patřících k frekvenční špičce mluvené, či psané češtiny. Za tu byl stanoven dataset 3000 nejčastějších lemmat v korpusech ORAL v1 a ORTOFON v1, resp. SYN2015; v rozdílech obou množin lze spatřovat projev diglosie v češtině na lexikální úrovni. Blíže analyzovány jsou tyto skupiny slov: adverbia, deiktika, numeralia, propria, deminutiva, přechýlená jména a univerbizáty.
[ "Michal Škrabal", "Zuzana Laubeová", "Barbora Štěpánková" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
AARGH! End-to-end Retrieval-Generation for Task-Oriented Dialog
AARGH! Kombinace vyhledání a generování v jednom modelu pro dialog zaměřený na úkoly
We introduce AARGH, an end-to-end task-oriented dialog system combining retrieval and generative approaches in a single model, aiming at improving dialog management and lexical diversity of outputs. The model features a new response selection method based on an action-aware training objective and a simplified single-encoder retrieval architecture which allow us to build an end-to-end retrieval-enhanced generation model where retrieval and generation share most of the parameters. On the MultiWOZ dataset, we show that our approach produces more diverse outputs while maintaining or improving state tracking and context-to-response generation performance, compared to state-of-the-art baselines.
Představujeme AARGH, komplexní dialogový systém orientovaný na úkoly, který kombinuje vyhledávací a generativní přístupy v jediném modelu a jehož cílem je zlepšit řízení dialogu a lexikální rozmanitost výstupů. Model obsahuje novou metodu výběru odpovědí založenou na trénovacím cíli zohledňujícím akce a zjednodušenou architekturu vyhledávání s jedním enkodérem, které nám umožňují vytvořit end-to-end model s generováním obohaceným o vyhledávání, kde vyhledávání a generování sdílí většinu parametrů. Na datové sadě MultiWOZ ukazujeme, že náš přístup produkuje rozmanitější výstupy při zachování nebo zlepšení výkonu sledování stavu a generování odpovědi z kontextu ve srovnání s nejmodernějšími referenčními modely. Translated with www.DeepL.com/Translator (free version)
[ "Tomáš Nekvinda", "Ondřej Dušek" ]
https://www.semanticscholar.org/paper/fd6d8564670f0c19cc32b14f4f705e37b0b9e423/
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
ELITR Demo at ECSPM Symposium
Ukázka projektu ELITR na Sympóziu ECSPM
I introduced the EU project ELITR, its goals and the current stage, and demoed it in practice.
Představil jsem projekt ELITR, jeho cíle i aktuální stav, a demonstroval jej v praxi.
[ "Ondřej Bojar" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
ELITR Overview at NPFL093: NLP Applications
Přehled o projektu ELITR pro seminář NPFL093
I presented the EU project ELITR at an invited talk of the seminar NPFL093 NLP Applications.
Na zvané přednášce v rámci semináře NPFL093 Aplikace NLP jsem představil EU projekt ELITR.
[ "Ondřej Bojar" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
Introduction to (Neural) Machine Translation
Úvod do (neuronového) strojového překladu
Opening talk of MT Marathon 2022, describing the task, approached and methods of evaluation.
Úvodní přednáška k týdenní letní škole MT Marathon 2022, popis úlohy překladu, přístupů i metod vyhodnocování.
[ "Ondřej Bojar" ]
https://www.semanticscholar.org/paper/405f1a5602867c66e015491c26d2be5504eed458/
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
Live Subtitling and First Steps towards Meeting Summarization
Živé titulkování a první kroky k automatickému dělání zápisků z jednání
I presented the EU project ELITR at an in-house seminar of the "Interinstitutional Task Force on Speech Recognition" of the European Commission.
Na zvaném interním semináři "pracovní skupiny pro rozpoznávání mluvené řeči" Evropské komise jsem představil projekt ELITR.
[ "Ondřej Bojar" ]
https://www.semanticscholar.org/paper/5253bc2e757012fa41af5f0d82672f3f3131d029/
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
Machine Translation Usable in Practice, Let's Move to Minuting
Strojový překlad použitelný v praxi, přesuňme se k zápiskování
I introduced the results of EU projects Bergamot and ELITR, and also the task of automatic minuting.
Představil jsem výsledky projektů Bergamot a ELITR a úlohu automatického tvoření zápisků z jednání.
[ "Ondřej Bojar" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
cs
2022
Machine translation and "machine interpreting": state of the art
Strojový překlad a strojové „tlumočenı́“: Co se aktuálně umí
An introduction of the task of machine translation and speech translation, including the current limitations.
Představení úlohy strojového překladu a strojového překladu mluvené řeči, včetně současných omezení.
[ "Ondřej Bojar" ]
https://www.semanticscholar.org/paper/f0225824c57b565431d0783df58356076c903d4f/
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
Two Extremes in Machine Translation
Dva extrémy se strojovém překladu
I presented the results of two EU projects, ELITR and Bergamot: highly multilingual translation and translation on personal computers.
Představil jsem výsledky dvou EU projetků, ELITR a Bergamot, jako dva extrémy v nasazení strojového překladu: Překlad do mnoha jazyků a překlad na osobním počítači.
[ "Ondřej Bojar" ]
https://www.semanticscholar.org/paper/63a3abb3b045c814e8c34864895f04e7e004362d/
null
null
null
null
null
null
null
null
null
null
null
null
cs
2022
LangTools for ERSTE
LangTools pro ERSTE
Prezentace jazykových nástrojů pro Českou spořitelnu. Rozpoznávání mluvené řeči, strojový překlad v rámci okamžité odpovědi, mezijazyčné vyhledávání informací, odpovídání na otázky, automatické shrnování schůzek.
Presentation of language tools for Česká spořitelna. Speech Recognition, Immediate Response Machine Translation, Cross-Lingual Information Retrieval, Question Answering, Automatic Summarization of Meetings.
[ "Ondřej Bojar", "Jindřich Libovický", "Rudolf Rosa", "Tomáš Musil", "Matúš Žilinec" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
LUSyD: From Syntax to DiscourseffNotes on Knowledge Representation (preliminary
LUSyD: Od Syntaxe k Diskursu, Reprezentace významu
LuSyd is a project about meaning representation using SynSemClass ontology. Multilingual (English, Czech, German, Spanish so far) classes of synonyms will be used to annotate meaning in the corpus.
LuSyd je projekt o reprezentaci významu s využitím ontologie SynSemClass. Vícejazyčné (zatím angličtina, čeština, němčina, španělština) třídy synonym budou použity k anotaci významu v korpusu.
[ "Jan Hajič" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
LINDAT/CLARIAH-CZ: Where We Are and Where We Go
LINDAT/CLARIAH-CZ: Kde jsme a kam míříme
In this chapter we present the main achievements of the Czech large research infrastructure LINDAT/CLARIAH-CZ. We provide a short description of the infrastructure and its history, and a brief account of its scientific, technological and infrastructural scope. We focus on the technological innovations already implemented in the repository and in the service offerings, and outline some future plans.
V této kapitole uvádíme hlavní úspěchy české velké výzkumné infrastruktury LINDAT/CLARIAH-CZ. Poskytujeme krátký popis infrastruktury a její historii a stručný popis jejího vědeckého, technologického a infrastrukturního rozsahu. Zaměřujeme se na technologické inovace již implementované v úložišti a v nabídkách služeb a nastiňujeme některé budoucí plány.
[ "Jan Hajič", "Eva Hajičová", "Barbora Hladká", "Ondřej Košarko", "Jozef Mišutka", "Pavel Straňák" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
Multilingual SynSemClass for the Semantic Web (MSSW)
Vícejazyčný SynSemClass pro sémantický web (MSSW)
LLOD (Linguistic Linked Open Data) is a generic name for a set of mutually connected language resources, using ontological relations. The connections between concepts and between concepts and their expressions in natural language make them suitable for both research and industrial applications in the area of content analysis, natural language understanding, (language- and knowledge-based) inferencing and other tasks. In the presented task, the concrete work will be on converting the SynSemClass project dataset (in part as a result of a previous Humane AI Net microproject called META-O-NLU) into LLOD, connecting it to the huge amount or interlinked data already available. A partner is involved in the Prêt-à-LLOD H2020 project, making this project synergistic in nature and multiplicative in terms of results in previous projects. Partners are also involved in the COST Action “European network for Web-centered linguistic data science” (NexusLinguarum).
LLOD (Linguistic Linked Open Data) je obecný název pro soubor vzájemně propojených jazykových zdrojů, využívajících ontologické vztahy. Propojení mezi pojmy a mezi pojmy a jejich výrazy v přirozeném jazyce je činí vhodnými jak pro výzkumné, tak pro průmyslové aplikace v oblasti analýzy obsahu, porozumění přirozenému jazyku, (na jazyku a znalostech založených) odvozování a dalších úloh. V předloženém úkolu bude konkrétní práce na převedení datového souboru projektu SynSemClass (částečně jako výsledek předchozího mikroprojektu Humane AI Net s názvem META-O-NLU) do LLOD a jeho propojení s obrovským množstvím nebo vzájemně propojenými daty, která jsou již k dispozici. Partner je zapojen do projektu Prêt-à-LLOD H2020, díky čemuž má tento projekt synergický charakter a multiplikativní charakter, pokud jde o výsledky předchozích projektů. Partneři jsou také zapojeni do akce COST „Evropská síť pro Web-centered linguistic data science“ (NexusLinguarum).
[ "Jan Hajič", "Zdeňka Urešová", "Eva Fučíková", "Cristina Fernández-Alcaina", "Jana Straková", "Thierry Declerck" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
cs
2022
Synonymy and homonymy in grammar
Synonymie a homonymie v gramatice
In this paper examples of synonymy and homonymy in Czech are presented. The sources of homonymy are exemplified by the morphological, syntactic and lexical means and their combinations in the sentence leading to their surface identity. Synonymy is demonstrated by sentences with different syntactic and morphemic forms expressing the same meaning. Local adverbials with the meaning Where? (LOC) are analyzed here in a more detail. Attention is focused on the analysis of the prepositional constructions with the preposition v(e) + Loc [in] and na + Loc [on] and the differences between the construction including them as well as examples of their interchangeability are demonstrated. Prepositional constructions with the preposition u + Gen [at/by] are added to the central domain of the local adverbials connected with the answer to the question Where? [LOC]. Criteria for the description of grammatical requirements for their form as well as for their lexicalized forms applied in this domain are proposed as the sources of the relation of homonymy and synonymy. The analysis is, however, limited to the narrow part of local meanings with the aim to present both sides of the language signs from the point of view of their asymmetry demonstrated by means of the transparent part of grammatical description.
V tomto příspěvku jsou uvedeny příklady synonymie a homonymie v češtině. Zdrojem homonymie jsou morfologické, syntaktické a lexikální prostředky a jejich kombinace ve větě vedoucí k jejich povrchové identitě. Synonymii demonstrují věty s různými syntaktickými a morfemickými formami vyjadřujícími stejný význam. Místní příslovce s významem Kde? (LOC) jsou zde analyzovány podrobněji. Pozornost je zaměřena na analýzu předložkových konstrukcí s předložkou v(e) + Loc [in] a na + Loc [on], u + Gen [at/by]. Jsou navržena kritéria pro určování synonymie a gramatických požadavků na formu lokálních určení. Analýza se omezuje na úzkou část lokálních významů s cílem představit obě strany jazykového znaku z hlediska jeho asymetrie.
[ "Jarmila Panevová", "Marie Mikulová" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
SiR 1.0
SiR 1.0
The SiR 1.0 corpus is a collection of articles published on the iRozhlas server with a manual annotation of citations. The annotators marked and linked citation phrases with citation sources and decided on a type of the source. The corpus consists of 1 718 articles, 589 of them were double- or triple-annotated.
Korpus SiR 1.0 je kolekce článků publikovaných na serveru iRozhlas s ruční anotací citací. Anotátoři označovali a propojovali citační fráze a zdroje a určovali typ zdroje. Korpus obsahuje celkem 1 718 článků, 589 z nich anotovaných dvojitě či trojitě.
[ "Barbora Vidová Hladká", "Jiří Mírovský", "Matyáš Kopp", "Václav Moravec" ]
https://www.semanticscholar.org/paper/f6ae446b43f9cdc22da4960835aec49e828cac25/
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
Towards Universal Segmentations: UniSegments 1.0
K Universal Segmentations: UniSegments 1.0
Our work aims at developing a multilingual data resource for morphological segmentation. We present a survey of 17 existing data resources relevant for segmentation in 32 languages, and analyze diversity of how individual linguistic phenomena are captured across them. Inspired by the success of Universal Dependencies, we propose a harmonized scheme for segmentation representation, and convert the data from the studied resources into this common scheme. Harmonized versions of resources available under free licenses are published as a collection called UniSegments 1.0.
Náš článek se zaměřuje na vývoj vícejazyčného zdroje dat pro morfologickou segmentaci. Představujeme přehled 17 existujících datových zdrojů relevantních pro segmentaci ve 32 jazycích a analyzujeme rozmanitost způsobů, jakými jsou v nich jednotlivé jazykové jevy zachyceny. Nechali jsme se inspirovat úspěchem Universal Dependencies a navrhujeme harmonizované schéma pro reprezentaci segmentaci a převádíme data z těchto zdrojů do jednotného schématu. Harmonizované verze zdrojů dostupné pod bezplatnými licencemi jsou publikovány jako kolekce s názvem UniSegments 1.0.
[ "Zdeněk Žabokrtský", "Niyati Bafna", "Jan Bodnár", "Lukáš Kyjánek", "Emil Svoboda", "Magda Ševčíková", "Jonáš Vidra" ]
https://www.semanticscholar.org/paper/c3103c6d21b6d11feb1011e9674be50789c8ae3e/
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
Universal Segmentations 1.0 (UniSegments 1.0)
Universal Segmentations 1.0 (UniSegments 1.0)
Universal Segmentations (UniSegments) is a collection of lexical resources capturing morphological segmentations harmonised into a cross-linguistically consistent annotation scheme for many languages. The annotation scheme consists of simple tab-separated columns that stores a word and its morphological segmentations, including pieces of information about the word and the segmented units, e.g., part-of-speech categories, type of morphs/morphemes etc. The current public version of the collection contains 38 harmonised segmentation datasets covering 30 different languages.
Universal Segmentations (UniSegments) je kolekce lexikálních zdrojů zachycujících morfologické segmentace mnoha jazyků harmonizované do lingvisticky konzistentního anotačního schématu. Anotační schéma je uloženo jednoduchém sloupcovém formátu, přičemž jednotlivé sloupce jsou odděleny tabulátory. K jednotlivým slovům se ukládá jejich morfologická segmentace, včetně různých informací o slovech a segmentovaných jednotkách, např. slovní druhy, typy morfů/morfémů atd. Současná veřejná verze kolekce obsahuje 38 harmonizovaných datových souborů pokrývajících 30 různých jazyků.
[ "Zdeněk Žabokrtský", "Niyati Bafna", "Jan Bodnár", "Lukáš Kyjánek", "Emil Svoboda", "Magda Ševčíková", "Jonáš Vidra", "Sachi Angle", "Ebrahim Ansari", "Timofey Arkhangelskiy", "Khuyagbaatar Batsuren", "Gábor Bella", "Pier Marco Bertinetto", "Olivier Bonami", "Chiara Celata", "Michael Daniel", "Alexei Fedorenko", "Matea Filko", "Fausto Giunchiglia", "Hamid Haghdoost", "Nabil Hathout", "Irina Khomchenkova", "Victoria Khurshudyan", "Dmitri Levonian", "Eleonora Litta", "Maria Medvedeva", "S. N. Muralikrishna", "Fiametta Namer", "Mahshid Nikravesh", "Sebastian Padó", "Marco Passarotti", "Vladimir Plungian", "Alexey Polyakov", "Mihail Potapov", "Mishra Pruthwik", "Ashwath B. Rao", "Sergei Rubakov", "Husain Samar", "Dipti Misra Sharma", "Jan Šnajder", "Krešimir Šojat", "Vanja Štefanec", "Luigi Talamo", "Delphine Tribout", "Daniil Vodolazsky", "Arseniy Vydrin", "Aigul Zakirova", "Britta Zeller" ]
https://www.semanticscholar.org/paper/c3103c6d21b6d11feb1011e9674be50789c8ae3e/
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
Findings of the Shared Task on Multilingual Coreference Resolution
Výsledky soutěže ve vícejazyčné automatické analýze koreference
This paper presents an overview of the shared task on multilingual coreference resolution associated with the CRAC 2022 workshop. Shared task participants were supposed to develop trainable systems capable of identifying mentions and of clustering the mentions according to identity coreference. The public edition of CorefUD 1.0, which contains 13 datasets for 10 languages was used as the source of training and evaluation data. The CoNLL score used in previous coreference-oriented shared tasks was used as the main evaluation metric. There were 8 coreference prediction systems submitted by 5 participating teams; in addition, there was a competitive transformer-based baseline system provided by the organizers at the beginning of the shared task. The winner system outperformed the baseline by 12 percent points (in terms of the CoNLL scores averaged across all datasets for individual languages).
Tento článek představuje přehled výsledků soutěže (společné úlohy) ve vícejazyčné automatické analýze koreference, která byla přidružená k workshopu CRAC 2022. Účastníci soutěže měli vyvinout trénovatelné systémy schopné identifikovat zmínky o entitách a shlukovat tyto zmínky na základě identické koreference. Jako zdroj trénovacích a vyhodnocovacích dat byla použita veřejná část CorefUD 1.0, která obsahuje 13 korpusů pro 10 jazyků. Jako hlavní vyhodnocovací metrika bylo použito skóre CoNLL, které se používalo v dřívějších soutěžích zaměřených na koreferenci. 5 účastnických týmů vyvinulo celkem 8 systémů na predikci koreference; kromě toho jsou k dispozici výsledky baseline systému, který je založen na transformerech a byl poskytnut organizátory na začátku soutěže. Vítězný systém překonal baseline o 12 procentních bodů (průměr CoNLL skóre přes korpusy jednotlivých jazyků).
[ "Zdeněk Žabokrtský", "Miloslav Konopík", "Anna Nedoluzhko", "Michal Novák", "Maciej Ogrodniczuk", "Martin Popel", "Ondřej Pražák", "Jakub Sido", "Daniel Zeman", "Yilun Zhu" ]
https://www.semanticscholar.org/paper/cd23e53717e9ab628c8bd689c57deff74bd639e8/
null
null
null
null
null
null
null
null
null
null
null
null
cs
2022
From a lingustic theory towards an annotated corpus and back
Cesta od lingvistické teorie k anotovanému korpusu a zpátky
A theoretically well-founded annotation of language corpora is a very importatnt component part of such corpora. In our paper, we support this point by the example of the Prague Dependency Treebank and its annotation scheme. We bring some case studies concerning the information structure of the sentence. At the same time, we also demonstrate that the work with a consistently annotated corpus may lead to some modifications of the original theory.
Dobře teoreticky podložená všestranná anotace je velmi důležitou součástí jazykových korpusů. Jako příklad ve stati uvádíme Pražský závislostní korpus, který obsahuje anotace na několika jazykových rovinách, a tento přístup ilustrujeme na několika případkových studiích z oblasti aktuálního členění. Zároveň na těchto příkladech dokládáme, že práce s anotovaným korpusem může vést k upřesnění výchozí teorie.
[ "Eva Hajičová" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
cs
2022
Patrice Pognan (non)calculable
Patrice Pognan (ne)vypočítatelný
A glimpse at an effort to explain certain regularities as well as (seemingly) irregularities in Czech and other Slavonic languages (first of all in the domain of inflectional and derivational morphology) by means of rules, i.e. to "calculate" these phenomena.
Zamyšlení nad snahou vysvětlit některé pravidelnosti a zdánlivé nepravidelnosti v češtině a jiných slovanských jazycích (především v oblasti deklinační i slovotvorné morfologie) pomocí pravidel, tedy tyto jevy "vypočítat".
[ "Eva Hajičová" ]
https://www.semanticscholar.org/paper/1f8b9c2c29a61d2795a07c3a5f4c87ac22be29aa/
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
Corpus Annotation as a Feasible and Scientifically Beneficial Task
Korpusová anotace jako proveditelný a vědecky prospěšný úkol
The aim of the present chapter is to demonstrate that a well-designed and theoretically founded corpus annotation contributes significantly to the use of the corpus for testing a linguistic theory and its further development. The data for our analyses come from the Prague Dependency Treebank family both monolingual Czech and parallel English-Czech and concern the underlying syntactic level of language description and the annotation of discourse structure. In particular, the case studies concern three research questions, namely (i) the semantic relevance of information structure of the sentence, (ii) the relation between focus sensitive particles and discourse connectives with respect to the semantics of discourse relations, and (iii) the relation between primary and secondary connectives. In the Appendix, some data on measuring inter-annotator agreement are presented and discussed.
Cílem této kapitoly je ukázat, že dobře navržená a teoreticky podložená korpusová anotace významně přispívá k využití korpusu pro testování lingvistické teorie a pro její další rozvoj. Data našich analýz pocházejí z korpusové rodiny Prague Dependency Treebank; jde o jednojazyčná česká data a o paralelní anglicko-česká data, která se týkají základní syntaktické úrovně popisu jazyka a anotace struktury diskurzu. Konkrétní případové studie se týkají tří výzkumných otázek, a to (i) sémantické relevance informační struktury věty, (ii) vztahu mezi fokalizátory a diskurzními konektory vzhledem k sémantice diskurzních vztahů a (iii) vztahu mezi primárními a sekundárními spojovacími výrazy. V příloze jsou uvedeny a rozebrány údaje o měření mezianotátorské shody.
[ "Eva Hajičová", "Jan Hajič", "Barbora Hladká", "Jiří Mírovský", "Lucie Poláková", "Kateřina Rysová", "Magdaléna Rysová", "Pavel Straňák", "Barbora Štěpánková", "Šárka Zikánová" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
Information structure in a formal description of language as reflected in an annotated corpus of Czech
Informační struktura ve formálním popisu jazyka, jak se odráží v anotovaném korpusu češtiny
A brief comparison of the theoretical approach to the information structure in the Meaning-Text Theory (MTT) and in the Functional Generative Description (FGD) is followed by an account of the treatment of the given phenomenon in an annotated corpus of running texts. The FGD approach is documented by several case studies demonstrating how an annotated corpus helps to verify and to define with more precision linguistic hypotheses.
Po krátkém srovnání teoretického přístupu k informační struktuře v Meaning-Text Theory (MTT) a ve Funkčním generativním popisu (FGD) následuje popis zpracování daného jevu v anotovaném korpusu reálných textů. Přístup FGD je dokumentován několika případovými studiemi demonstrujícími, jak anotovaný korpus pomáhá ověřovat a přesněji definovat lingvistické hypotézy.
[ "Eva Hajičová", "Marie Mikulová" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
Advantages of a complex multilayer annotation scheme: The case of the Prague Dependency Treebank
Výhody komplexního více-rovinného anotačního schématu: Případ Pražského závislostního korpusu
Recently, many corpora have been developed that contain multiple annotations of various linguistic phenomena, from morphological categories of words through the syntactic structure of sentences to discourse and coreference relations in texts. Discussions are ongoing on an appropriate annotation scheme for a large amount of diverse information. In our contribution we express our conviction that a multilayer annotation scheme offers to view the language system in its complexity and in the interaction of individual phenomena and there are at least two aspects that support a multilayer annotation scheme: (i) A multilayer annotation scheme makes it possible to use the annotation of one layer to design the annotation of another layer(s) both conceptually and in a form of a pre-annotation procedure or annotation checking rules. (ii) A multilayer annotation scheme presents a reliable ground for corpus studies based on features across the layers. These aspects are demonstrated on the case of the Prague Dependency Treebank. Its multilayer annotation scheme withstood the test of time and serves well also for complex textual annotations, in which earlier morpho-syntactic annotations are advantageously used. In addition to a reference to the previous projects that utilise its annotation scheme, we present several current investigations.
V poslední době bylo vyvinuto mnoho korpusů, které obsahují více anotací různých jazykových jevů, od morfologických kategorií slov přes syntaktickou stavbu vět až po diskurz a koreferenční vztahy v textech. Probíhají diskuse o vhodném anotačním schématu pro velké množství různorodých informací. V našem příspěvku vyjadřujeme přesvědčení, že vícevrstvé anotační schéma nabízí pohled na jazykový systém v jeho komplexnosti a v interakci jednotlivých jevů a že existují minimálně dva aspekty, které vícevrstvé anotační schéma podporují: (i) Vícevrstvé anotační schéma umožňuje použít anotaci jedné vrstvy k návrhu anotace další vrstvy (vrstev) jak koncepčně, tak ve formě předanotačního postupu nebo pravidel kontroly anotace. (ii) Vícevrstvé anotační schéma představuje spolehlivý základ pro korpusové studie založené na vlastnostech napříč vrstvami. Tyto aspekty jsou demonstrovány na případu Pražského závislostního korpusu. Jeho více-rovinné anotační schéma obstálo ve zkoušce času a dobře poslouží i pro složité textové anotace, ve kterých se s výhodou používají dřívější morfosyntaktické anotace. Kromě odkazu na předchozí projekty, které využívají toto anotační schéma, uvádíme několik aktuálních výzkumů.
[ "Eva Hajičová", "Marie Mikulová", "Barbora Štěpánková", "Jiří Mírovský" ]
https://www.semanticscholar.org/paper/c2e9552a6d3f683469e547cfba0841ff465c0c50/
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
PDT-Vallex: Valency Lexicon Linked to Czech Corpora
PDT-Vallex: valenční slovník propojený s českými korpusy
The paper is about the project of the PDT-Vallex valence dictionary, which was created together with the PDT annotations and was then further expanded and used to annotate the corpus.
Příspěvek je o projektu valenčního slovníku PDT-Vallex, který vznikal spolu s anotacemi PDT a pak byl dále rozšiřován a používán k anotacím korpusu.
[ "Zdeňka Urešová" ]
https://www.semanticscholar.org/paper/39b9394be0cf60a671d201a8074ccbe473ee93b5/
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
SynSemClass - an Event-type Ontology in Multiple Languages
SynSemClass- ontologie dějů pro více jazyků
SynSemClass is an extension of the bilingual Czech-English dictionary of verbal synonyms including German and Spanish synonyms. It brings together verbal synonyms according to their meanings into classes that contain their semantic and valency characteristics.
Slovník SynSemClass je rozšířením dvojjazyčného česko-anglického slovníku slovesných synonym o německá a španělská synonyma. Sdružuje slovesné synonymní děje podle významu do tříd, které obsahují jejich významovou a valenční charakteristiku.
[ "Zdeňka Urešová" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
SynSemClass 4.0
SynSemClass 4.0
The SynSemClass synonym verb lexicon version 3.5 investigates, with respect to contextually-based verb synonymy, semantic ‘equivalence’ of Czech, English, and German verb senses and their valency behavior in parallel Czech-English and German-English language resources. SynSemClass4.0 is a multilingual event-type ontology based on classes of synonymous verb senses, complemented with semantic roles and links to existing semantic lexicons. The SynSemClass is not only enriched by an additional number of classes but in the context of content hierarchy, some classes have been merged. Compared to the older version of the lexicon, the novelty is the definitions of classes and the definitions of roles. Apart from the already used links to PDT-Vallex, EngVallex, CzEngVallex, FrameNet, VerbNet, PropBank, Ontonotes, and English WordNet for Czech and English entries the new links to German language lexical resources are exploited for German verb entries, such as Woxikon, E-VALBU, and GUP. The German part of the lexicon has been created within the project Multilingual Event-Type-Anchored Ontology for Natural Language Understanding (META-O-NLU) by two cooperating teams - the team of the Charles University, Faculty of Mathematics and Physics, Institute of Formal and Applied Linguistics, Prague (ÚFAL), Czech Republic and the team of the German Research Center for Artificial Intelligence (DFKI) Speech and Language Technology, Berlin, Germany.
SynSemClass lexicon 4.0 zkoumá s ohledem na kontextově založenou slovesnou synonymii, sémantickou „ekvivalenci“ českých, anglických a německých sloves a jejich valenční chování v paralelních česko-anglických a německo-anglických jazykových zdrojích. SynSemClass 4.0 je vícejazyčná ontologie typu události založená na třídách synonymních významů sloves, doplněná sémantickými rolemi a odkazy na existující sémantické lexikony. Slovník je nejen obohacen o další množství tříd, ale v rámci hierarchizace obsahu byly některé třídy sloučeny. V porovnání se starou verzí obsahuje slovník nově definice tříd a definice rolí. Kromě již použitých odkazů na položky PDT-Vallex, EngVallex, CzEngVallex, FrameNet, VerbNet, PropBank, Ontonotes a English WordNet pro české a anglické záznamy jsou nové odkazy na německé jazykové lexikální zdroje, jako je Woxikon, E-VALBU a GUP, využívány pro německé slovesné záznamy. Německá část lexikonu byla vytvořena v rámci projektu Multilingual Event-Type-Anchored Ontology for Natural Language Understanding (META-O-NLU) dvěma spolupracujícími týmy - týmem Univerzity Karlovy, Matematicko-fyzikální fakulty, Ústavu formální a aplikované lingvistiky, Praha (ÚFAL), Česká republika a týmem Německého výzkumného centra pro umělou inteligenci (DFKI) Speech and Language Technology, Berlín, Německo.
[ "Zdeňka Urešová", "Peter Bourgonje", "Eva Fučíková", "Jan Hajič", "Eva Hajičová", "Kateřina Rysová", "Georg Rehm", "Karolina Zaczynska" ]
https://www.semanticscholar.org/paper/6cb63773f1924add840f59357e988cf2481672a6/
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
Making a Semantic Event-type Ontology Multilingual
Zvícejazyčnění sémantické ontologie
We present an extension of the SynSemClass event-type ontology, originally conceived as a bilingual Czech-English resource. We added German entries to the classes representing the concepts of the ontology. Having a different starting point than the original work (unannotated parallel corpus without links to a valency lexicon and, of course, different existing lexical resources), it was a challenge to adapt the annotation guidelines, the data model and the tools used for the original version. We describe the process and results of working in such a setup. We also show the next steps to adapt the annotation process, data structures and formats and tools necessary to make the addition of a new language in the future more smooth and efficient, and possibly to allow for various teams to work on SynSemClass extensions to many languages concurrently. We also present the latest release which contains the results of adding German, freely available for download as well as for online access.
Představujeme rozšíření ontologie typu události SynSemClass, původně koncipované jako dvojjazyčný česko-anglický zdroj. Do tříd představujících koncepty ontologie jsme přidali německé záznamy. Vzhledem k tomu, že jsme měli jiný výchozí bod než původní práce (neoznačený paralelní korpus bez odkazů na valenční lexikon a samozřejmě různé existující lexikální zdroje), bylo náročné přizpůsobit pokyny pro anotaci, datový model a nástroje použité pro původní verzi. Popisujeme proces a výsledky práce v takovém nastavení. Dále ukazujeme další kroky k úpravě procesu anotace, datové struktury a formáty a nástroje nezbytné k tomu, aby přidání nového jazyka v budoucnu bylo plynulejší a efektivnější a případně aby různé týmy mohly pracovat na rozšíření SynSemClass do mnoha jazyků současně. představujeme nejnovější verzi, která obsahuje výsledky přidání němčiny, volně dostupné ke stažení i pro online přístup.
[ "Zdeňka Urešová", "Karolina Zaczynska", "Peter Bourgonje", "Eva Fučíková", "Georg Rehm", "Jan Hajič" ]
https://www.semanticscholar.org/paper/d85cf7ddec7872c8c9f69288e1b30eea5b0a6061/
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
NomVallex: A Valency Lexicon of Czech Nouns and Adjectives
NomVallex: Valenční slovník českých substantiv a adjektiv
We present NomVallex, a manually annotated valency lexicon of Czech nouns and adjectives. The lexicon is created in the theoretical framework of the Functional Generative Description and based on corpus data. In total, NomVallex 2.0 is comprised of 1027 lexical units contained in 570 lexemes, covering the following part-of-speech and derivational categories: deverbal and deadjectival nouns, and deverbal, denominal, deadjectival and primary adjectives. Valency properties of a lexical unit are captured in a valency frame which is modeled as a sequence of valency slots, supplemented with a list of morphemic forms. In order to make it possible to study the relationship between valency behavior of base words and their derivatives, lexical units of nouns and adjectives in NomVallex are linked to their respective base words, contained either in NomVallex itself or, in case of verbs, in a valency lexicon of Czech verbs called VALLEX. NomVallex enables a comparison of valency properties of a significant number of Czech nominals with their base words, both manually and in an automatic way; as such, we can address the theoretical question of argument inheritance, concentrating on systemic and non-systemic valency behavior.
Představujeme NomVallex, ručně anotovaný valenční slovník českých substantiv a adjektiv. Slovník je vytvořený v teoretickém rámci Funkčního generativního popisu a je založený na korpusových datech. Obsahuje celkem 1027 lexikálních jednotek v celkovém počtu 570 lexémů, přičemž zahrnuje následující slovnědruhové a derivační kategorie: deverbální a deadjektivní substantiva, a deverbální, desubstantivní, deadjektivní a primární adjektiva. Valenční vlastnosti lexikálních jednotek zachycuje v podobě valenčního rámce, který pro každé valenční doplnění uvádí funktor a množinu možných morfematických vyjádření, a dokládá je pomocí příkladů, které se vyskytly v použitých korpusech. NomVallex je koncipován jako lexikografický zdroj umožňující výzkum valence derivačně příbuzných lexikálních jednotek, proto v relevantních případech poskytuje odkaz od určité lexikální jednotky k odpovídající lexikální jednotce jejího základového slova, obsaženého buď v NomVallexu, nebo – v případě sloves – ve slovníku VALLEX, čímž propojuje až tři slovní druhy, konkrétně substantivum – sloveso, adjektivum – sloveso, substantivum – adjektivum nebo substantivum – adjektivum – sloveso. NomVallex umožňuje srovnání valenčního chování velkého počtu českých substantiv a adjektiv s valencí jejich základových slov, což je jeden z předpokladů ke zkoumání teoretické otázky dědičnosti valence a k popisu systémového a nesystémového valenčního chování.
[ "Veronika Kolářová", "Anna Vernerová" ]
https://www.semanticscholar.org/paper/6939d3c0797d6b1967d05a90b23e1cd74029df36/
null
null
null
null
null
null
null
null
null
null
null
null
cs
2022
NomVallex 2.0
NomVallex 2.0
NomVallex 2.0 is a manually annotated valency lexicon of Czech nouns and adjectives, created in the theoretical framework of the Functional Generative Description and based on corpus data (the SYN series of corpora from the Czech National Corpus and the Araneum Bohemicum Maximum corpus). In total, NomVallex is comprised of 1027 lexical units contained in 570 lexemes, covering the following parts-of-speech and derivational categories: deverbal or deadjectival nouns, and deverbal, denominal, deadjectival or primary adjectives. Valency properties of a lexical unit are captured in a valency frame (modeled as a sequence of valency slots, each supplemented with a list of morphemic forms) and documented by corpus examples. In order to make it possible to study the relationship between valency behavior of base words and their derivatives, lexical units of nouns and adjectives in NomVallex are linked to their respective base lexical units (contained either in NomVallex itself or, in case of verbs, in the VALLEX lexicon), linking up to three parts-of-speech (i.e., noun – verb, adjective – verb, noun – adjective, and noun – adjective – verb). In order to facilitate comparison, this submission also contains abbreviated entries of the base verbs of these nouns and adjectives from the VALLEX lexicon and simplified entries of the covered nouns and adjectives from the PDT-Vallex lexicon.
NomVallex 2.0 je ručně anotovaný valenční slovník českých substantiv a adjektiv, vytvořený v teoretickém rámci Funkčního generativního popisu a založený na korpusových datech (korpusy řady SYN Českého národního korpusu a korpus Araneum Bohemicum Maximum). Slovník obsahuje celkem 1027 lexikálních jednotek v celkovém počtu 570 lexémů, přičemž zahrnuje následující slovnědruhové a derivační kategorie: (i) deverbální a deadjektivní substantiva, (ii) deverbální, desubstantivní, deadjektivní a primární adjektiva. Valenční vlastnosti lexikálních jednotek zachycuje v podobě valenčního rámce, který pro každé valenční doplnění uvádí funktor a množinu možných morfematických vyjádření, a dokládá je pomocí příkladů, které se vyskytly v použitých korpusech. NomVallex je koncipován jako lexikografický zdroj umožňující výzkum valence derivačně příbuzných lexikálních jednotek, proto v relevantních případech poskytuje odkaz od určité lexikální jednotky k odpovídající lexikální jednotce jejího základového slova, obsaženého buď v NomVallexu, nebo – v případě sloves – ve slovníku VALLEX, čímž propojuje až tři slovní druhy, konkrétně substantivum – sloveso, adjektivum – sloveso, substantivum – adjektivum nebo substantivum – adjektivum – sloveso. Za účelem umožnit srovnání obsahuje toto vydání rovněž zkrácená hesla základových sloves daných substantiv a adjektiv ze slovníku VALLEX a zjednodušená hesla příslušných substantiv a adjektiv ze slovníku PDT-Vallex.
[ "Veronika Kolářová", "Anna Vernerová", "Jana Klímová" ]
https://www.semanticscholar.org/paper/6939d3c0797d6b1967d05a90b23e1cd74029df36/
null
null
null
null
null
null
null
null
null
null
null
null
cs
2022
Parsing Old Czech with Modern Czech Models
Parsing staré češtiny s využitím modelů natrénovaných na nové češtině
In the first part of the talk we will present the Universal Dependencies project. In the second part we will focus on parsers trained on the Czech corpora in UD and their usage in building an Old Czech treebank.
V první části přednášky představíme projekt Universal Dependencies. Ve druhé se zaměříme na parsery natrénované na českých korpusech v UD a na jejich využití při budování staročeského treebanku.
[ "Daniel Zeman" ]
https://www.semanticscholar.org/paper/61097e5e330f320ff10ac66a2436da20d16bfdb6/
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
Universal Dependencies: Principles and Tools
Universal Dependencies: principy a nástroje
Universal Dependencies is an international community project and a collection of morphosyntactically annotated data sets (“treebanks”) for more than 100 languages. The collection is an invaluable resource for various linguistic studies, ranging from grammatical constructions within one language to language typology, documentation of endangered languages, and historical evolution of language. In the tutorial, I will first quickly show the main principles of UD, then I will present the actual data and various tools that are available to work with it: parsers, batch processors, search engines and viewers.
Universal Dependencies je mezinárodní komunitní projekt a sbírka morfosyntakticky anotovaných dat (treebanků) pro více než 100 jazyků. Tato sbírka je neocenitelným zdrojem pro různé jazykovědné studie od gramatických konstrukcí v jednom jazyku po jazykovou typologii, dokumentaci ohrožených jazyků a historického vývoje jazyka. V tomto tutoriálu nejdříve rychle představím hlavní principy UD, potom ukážu vlastní data a různé nástroje, které jsou k dispozici pro práci s nimi: parsery, dávkové procesory, vyhledávače a prohlížeče.
[ "Daniel Zeman" ]
https://www.semanticscholar.org/paper/3dcaf5e960a0d79df06a8cb1d603ad4555119db4/
null
null
null
null
null
null
null
null
null
null
null
null
cs
2022
Processing of Old Czech using Modern Czech Models
Zpracování staré češtiny s novočeskými modely
In the first part of the talk we will present the Universal Dependencies project. In the second part we will focus on parsers trained on the Czech corpora in UD and their usage in building an Old Czech treebank.
V první části přednášky představíme projekt Universal Dependencies. Ve druhé se zaměříme na parsery natrénované na českých korpusech v UD a na jejich využití při budování staročeského treebanku.
[ "Daniel Zeman" ]
https://www.semanticscholar.org/paper/b9b297e65866d746188aae8b785af995e7e0c26a/
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
Universal Dependencies 2.10
Universal Dependencies 2.10
Universal Dependencies is a project that seeks to develop cross-linguistically consistent treebank annotation for many languages, with the goal of facilitating multilingual parser development, cross-lingual learning, and parsing research from a language typology perspective. The annotation scheme is based on (universal) Stanford dependencies (de Marneffe et al., 2006, 2008, 2014), Google universal part-of-speech tags (Petrov et al., 2012), and the Interset interlingua for morphosyntactic tagsets (Zeman, 2008). This is the sixteenth release of UD Treebanks, Version 2.10.
Universal Dependencies je projekt, který vyvíjí mezijazykově konzistentní morfologicko-syntaktickou anotaci mnoha jazyků s cílem usnadnit vývoj jazykově nezávislých parserů, mezijazykové strojové učení a výzkum parsingu z perspektivy jazykové typologie. Anotační schéma je založeno na univerzálních Stanfordských závislostech (de Marneffe et al., 2006, 2008, 2014), univerzálních značkách Google (Petrov et al., 2012) a na Intersetu, tj. interlingvě pro sady morfologických značek (Zeman, 2008). Toto je šestnácté vydání treebanků UD, verze 2.10.
[ "Daniel Zeman", "Joakim Nivre", "Mitchell Abrams", "Elia Ackermann", "Noëmi Aepli", "Hamid Aghaei", "Željko Agić", "Amir Ahmadi", "Lars Ahrenberg", "Chika Kennedy Ajede", "Gabrielė Aleksandravičiūtė", "Ika Alfina", "Avner Algom", "Erik Andersen", "Lene Antonsen", "Katya Aplonova", "Angelina Aquino", "Carolina Aragon", "Glyd Aranes", "Maria Jesus Aranzabe", "Bilge Nas Arıcan", "Þórunn Arnardóttir", "Gashaw Arutie", "Jessica Naraiswari Arwidarasti", "Masayuki Asahara", "Deniz Baran Aslan", "Cengiz Asmazoğlu", "Luma Ateyah", "Furkan Atmaca", "Mohammed Attia", "Aitziber Atutxa", "Liesbeth Augustinus", "Elena Badmaeva", "Keerthana Balasubramani", "Miguel Ballesteros", "Esha Banerjee", "Sebastian Bank", "Verginica Barbu Mititelu", "Starkaður Barkarson", "Rodolfo Basile", "Victoria Basmov", "Colin Batchelor", "John Bauer", "Seyyit Talha Bedir", "Kepa Bengoetxea", "Yifat Ben Moshe", "Gözde Berk", "Yevgeni Berzak", "Irshad Ahmad Bhat", "Riyaz Ahmad Bhat", "Erica Biagetti", "Eckhard Bick", "Agnė Bielinskienė", "Kristín Bjarnadóttir", "Rogier Blokland", "Victoria Bobicev", "Loïc Boizou", "Emanuel Borges Völker", "Carl Börstell", "Cristina Bosco", "Gosse Bouma", "Sam Bowman", "Adriane Boyd", "Anouck Braggaar", "Kristina Brokaitė", "Aljoscha Burchardt", "Marie Candito", "Bernard Caron", "Gauthier Caron", "Lauren Cassidy", "Tatiana Cavalcanti", "Gülşen Cebiroğlu Eryiğit", "Flavio Massimiliano Cecchini", "Giuseppe G. A. Celano", "Slavomír Čéplö", "Neslihan Cesur", "Savas Cetin", "Özlem Çetinoğlu", "Fabricio Chalub", "Shweta Chauhan", "Ethan Chi", "Taishi Chika", "Yongseok Cho", "Jinho Choi", "Jayeol Chun", "Juyeon Chung", "Alessandra T. Cignarella", "Silvie Cinková", "Aurélie Collomb", "Çağrı Çöltekin", "Miriam Connor", "Daniela Corbetta", "Marine Courtin", "Mihaela Cristescu", "Philemon Daniel", "Elizabeth Davidson", "Mathieu Dehouck", "Martina de Laurentiis", "Marie-Catherine de Marneffe", "Valeria de Paiva", "Mehmet Oguz Derin", "Elvis de Souza", "Arantza Diaz de Ilarraza", "Carly Dickerson", "Arawinda Dinakaramani", "Elisa Di Nuovo", "Bamba Dione", "Peter Dirix", "Kaja Dobrovoljc", "Timothy Dozat", "Kira Droganova", "Puneet Dwivedi", "Hanne Eckhoff", "Sandra Eiche", "Marhaba Eli", "Ali Elkahky", "Binyam Ephrem", "Olga Erina", "Tomaž Erjavec", "Aline Etienne", "Wograine Evelyn", "Sidney Facundes", "Richárd Farkas", "Federica Favero", "Jannatul Ferdaousi", "Marília Fernanda", "Hector Fernandez Alcalde", "Jennifer Foster", "Cláudia Freitas", "Kazunori Fujita", "Katarína Gajdošová", "Daniel Galbraith", "Federica Gamba", "Marcos Garcia", "Moa Gärdenfors", "Sebastian Garza", "Fabrício Ferraz Gerardi", "Kim Gerdes", "Filip Ginter", "Gustavo Godoy", "Iakes Goenaga", "Koldo Gojenola", "Memduh Gökırmak", "Yoav Goldberg", "Xavier Gómez Guinovart", "Berta González Saavedra", "Bernadeta Griciūtė", "Matias Grioni", "Loïc Grobol", "Normunds Grūzītis", "Bruno Guillaume", "Céline Guillot-Barbance", "Tunga Güngör", "Nizar Habash", "Hinrik Hafsteinsson", "Jan Hajič", "Jan Hajič, jr.", "Mika Hämäläinen", "Linh Hà Mỹ", "Na-Rae Han", "Muhammad Yudistira Hanifmuti", "Takahiro Harada", "Sam Hardwick", "Kim Harris", "Dag Haug", "Johannes Heinecke", "Oliver Hellwig", "Felix Hennig", "Barbora Hladká", "Jaroslava Hlaváčová", "Florinel Hociung", "Petter Hohle", "Jena Hwang", "Takumi Ikeda", "Anton Karl Ingason", "Radu Ion", "Elena Irimia", "Ọlájídé Ishola", "Kaoru Ito", "Siratun Jannat", "Tomáš Jelínek", "Apoorva Jha", "Anders Johannsen", "Hildur Jónsdóttir", "Fredrik Jørgensen", "Markus Juutinen", "Sarveswaran Kengatharaiyer", "Hüner Kaşıkara", "Andre Kåsen", "Nadezhda Kabaeva", "Sylvain Kahane", "Hiroshi Kanayama", "Jenna Kanerva", "Neslihan Kara", "Ritván Karahóǧa", "Boris Katz", "Tolga Kayadelen", "Jessica Kenney", "Václava Kettnerová", "Jesse Kirchner", "Elena Klementieva", "Elena Klyachko", "Arne Köhn", "Abdullatif Köksal", "Kamil Kopacewicz", "Timo Korkiakangas", "Mehmet Köse", "Natalia Kotsyba", "Jolanta Kovalevskaitė", "Simon Krek", "Parameswari Krishnamurthy", "Sandra Kübler", "Oğuzhan Kuyrukçu", "Aslı Kuzgun", "Sookyoung Kwak", "Veronika Laippala", "Lucia Lam", "Lorenzo Lambertino", "Tatiana Lando", "Septina Dian Larasati", "Alexei Lavrentiev", "John Lee", "Phương Lê Hồng", "Alessandro Lenci", "Saran Lertpradit", "Herman Leung", "Maria Levina", "Cheuk Ying Li", "Josie Li", "Keying Li", "Yuan Li", "KyungTae Lim", "Bruna Lima Padovani", "Krister Lindén", "Nikola Ljubešić", "Olga Loginova", "Stefano Lusito", "Andry Luthfi", "Mikko Luukko", "Olga Lyashevskaya", "Teresa Lynn", "Vivien Macketanz", "Menel Mahamdi", "Jean Maillard", "Aibek Makazhanov", "Michael Mandl", "Christopher Manning", "Ruli Manurung", "Büşra Marşan", "Cătălina Mărănduc", "David Mareček", "Katrin Marheinecke", "Stella Markantonatou", "Héctor Martínez Alonso", "Lorena Martín Rodríguez", "André Martins", "Jan Mašek", "Hiroshi Matsuda", "Yuji Matsumoto", "Alessandro Mazzei", "Ryan McDonald", "Sarah McGuinness", "Gustavo Mendonça", "Tatiana Merzhevich", "Niko Miekka", "Karina Mischenkova", "Margarita Misirpashayeva", "Anna Missilä", "Cătălin Mititelu", "Maria Mitrofan", "Yusuke Miyao", "AmirHossein Mojiri Foroushani", "Judit Molnár", "Amirsaeid Moloodi", "Simonetta Montemagni", "Amir More", "Laura Moreno Romero", "Giovanni Moretti", "Keiko Sophie Mori", "Shinsuke Mori", "Tomohiko Morioka", "Shigeki Moro", "Bjartur Mortensen", "Bohdan Moskalevskyi", "Kadri Muischnek", "Robert Munro", "Yugo Murawaki", "Kaili Müürisep", "Pinkey Nainwani", "Mariam Nakhlé", "Juan Ignacio Navarro Horñiacek", "Anna Nedoluzhko", "Gunta Nešpore-Bērzkalne", "Manuela Nevaci", "Lương Nguyễn Thị", "Huyền Nguyễn Thị Minh", "Yoshihiro Nikaido", "Vitaly Nikolaev", "Rattima Nitisaroj", "Alireza Nourian", "Hanna Nurmi", "Stina Ojala", "Atul Kr. Ojha", "Adédayọ̀ Olúòkun", "Mai Omura", "Emeka Onwuegbuzia", "Noam Ordan", "Petya Osenova", "Robert Östling", "Lilja Øvrelid", "Şaziye Betül Özateş", "Merve Özçelik", "Arzucan Özgür", "Balkız Öztürk Başaran", "Teresa Paccosi", "Alessio Palmero Aprosio", "Hyunji Hayley Park", "Niko Partanen", "Elena Pascual", "Marco Passarotti", "Agnieszka Patejuk", "Guilherme Paulino-Passos", "Giulia Pedonese", "Angelika Peljak-Łapińska", "Siyao Peng", "Cenel-Augusto Perez", "Natalia Perkova", "Guy Perrier", "Slav Petrov", "Daria Petrova", "Andrea Peverelli", "Jason Phelan", "Jussi Piitulainen", "Tommi A Pirinen", "Emily Pitler", "Barbara Plank", "Thierry Poibeau", "Larisa Ponomareva", "Martin Popel", "Lauma Pretkalniņa", "Sophie Prévost", "Prokopis Prokopidis", "Adam Przepiórkowski", "Tiina Puolakainen", "Sampo Pyysalo", "Peng Qi", "Andriela Rääbis", "Alexandre Rademaker", "Mizanur Rahoman", "Taraka Rama", "Loganathan Ramasamy", "Carlos Ramisch", "Fam Rashel", "Mohammad Sadegh Rasooli", "Vinit Ravishankar", "Livy Real", "Petru Rebeja", "Siva Reddy", "Mathilde Regnault", "Georg Rehm", "Ivan Riabov", "Michael Rießler", "Erika Rimkutė", "Larissa Rinaldi", "Laura Rituma", "Putri Rizqiyah", "Luisa Rocha", "Eiríkur Rögnvaldsson", "Mykhailo Romanenko", "Rudolf Rosa", "Valentin Roșca", "Davide Rovati", "Ben Rozonoyer", "Olga Rudina", "Jack Rueter", "Kristján Rúnarsson", "Shoval Sadde", "Pegah Safari", "Benoît Sagot", "Aleksi Sahala", "Shadi Saleh", "Alessio Salomoni", "Tanja Samardžić", "Stephanie Samson", "Manuela Sanguinetti", "Ezgi Sanıyar", "Dage Särg", "Baiba Saulīte", "Yanin Sawanakunanon", "Shefali Saxena", "Kevin Scannell", "Salvatore Scarlata", "Nathan Schneider", "Sebastian Schuster", "Lane Schwartz", "Djamé Seddah", "Wolfgang Seeker", "Mojgan Seraji", "Syeda Shahzadi", "Mo Shen", "Atsuko Shimada", "Hiroyuki Shirasu", "Yana Shishkina", "Muh Shohibussirri", "Dmitri Sitchinava", "Janine Siewert", "Einar Freyr Sigurðsson", "Aline Silveira", "Natalia Silveira", "Maria Simi", "Radu Simionescu", "Katalin Simkó", "Mária Šimková", "Kiril Simov", "Maria Skachedubova", "Aaron Smith", "Isabela Soares-Bastos", "Shafi Sourov", "Carolyn Spadine", "Rachele Sprugnoli", "Vivian Stamou", "Steinþór Steingrímsson", "Antonio Stella", "Milan Straka", "Emmett Strickland", "Jana Strnadová", "Alane Suhr", "Yogi Lesmana Sulestio", "Umut Sulubacak", "Shingo Suzuki", "Daniel Swanson", "Zsolt Szántó", "Chihiro Taguchi", "Dima Taji", "Yuta Takahashi", "Fabio Tamburini", "Mary Ann C. Tan", "Takaaki Tanaka", "Dipta Tanaya", "Mirko Tavoni", "Samson Tella", "Isabelle Tellier", "Marinella Testori", "Guillaume Thomas", "Sara Tonelli", "Liisi Torga", "Marsida Toska", "Trond Trosterud", "Anna Trukhina", "Reut Tsarfaty", "Utku Türk", "Francis Tyers", "Sumire Uematsu", "Roman Untilov", "Zdeňka Urešová", "Larraitz Uria", "Hans Uszkoreit", "Andrius Utka", "Elena Vagnoni", "Sowmya Vajjala", "Rob van der Goot", "Martine Vanhove", "Daniel van Niekerk", "Gertjan van Noord", "Viktor Varga", "Uliana Vedenina", "Eric Villemonte de la Clergerie", "Veronika Vincze", "Natalia Vlasova", "Aya Wakasa", "Joel C. Wallenberg", "Lars Wallin", "Abigail Walsh", "Jing Xian Wang", "Jonathan North Washington", "Maximilan Wendt", "Paul Widmer", "Shira Wigderson", "Sri Hartati Wijono", "Seyi Williams", "Mats Wirén", "Christian Wittern", "Tsegay Woldemariam", "Tak-sum Wong", "Alina Wróblewska", "Mary Yako", "Kayo Yamashita", "Naoki Yamazaki", "Chunxiao Yan", "Koichi Yasuoka", "Marat M. Yavrumyan", "Arife Betül Yenice", "Olcay Taner Yıldız", "Zhuoran Yu", "Arlisa Yuliawati", "Zdeněk Žabokrtský", "Shorouq Zahra", "Amir Zeldes", "He Zhou", "Hanzhi Zhu", "Anna Zhuravleva", "Rayan Ziane" ]
https://www.semanticscholar.org/paper/1e105a47662affe083eaf110e5590ad4b9f7894d/
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
Universal Dependencies 2.11
Universal Dependencies 2.11
Universal Dependencies is a project that seeks to develop cross-linguistically consistent treebank annotation for many languages, with the goal of facilitating multilingual parser development, cross-lingual learning, and parsing research from a language typology perspective. The annotation scheme is based on (universal) Stanford dependencies (de Marneffe et al., 2006, 2008, 2014), Google universal part-of-speech tags (Petrov et al., 2012), and the Interset interlingua for morphosyntactic tagsets (Zeman, 2008). This is the seventeenth release of UD Treebanks, Version 2.11.
Universal Dependencies je projekt, který vyvíjí mezijazykově konzistentní morfologicko-syntaktickou anotaci mnoha jazyků s cílem usnadnit vývoj jazykově nezávislých parserů, mezijazykové strojové učení a výzkum parsingu z perspektivy jazykové typologie. Anotační schéma je založeno na univerzálních Stanfordských závislostech (de Marneffe et al., 2006, 2008, 2014), univerzálních značkách Google (Petrov et al., 2012) a na Intersetu, tj. interlingvě pro sady morfologických značek (Zeman, 2008). Toto je sedmnácté vydání treebanků UD, verze 2.11.
[ "Daniel Zeman", "Joakim Nivre", "Mitchell Abrams", "Elia Ackermann", "Noëmi Aepli", "Hamid Aghaei", "Željko Agić", "Amir Ahmadi", "Lars Ahrenberg", "Chika Kennedy Ajede", "Salih Furkan Akkurt", "Gabrielė Aleksandravičiūtė", "Ika Alfina", "Avner Algom", "Chiara Alzetta", "Erik Andersen", "Lene Antonsen", "Katya Aplonova", "Angelina Aquino", "Carolina Aragon", "Glyd Aranes", "Maria Jesus Aranzabe", "Bilge Nas Arıcan", "Þórunn Arnardóttir", "Gashaw Arutie", "Jessica Naraiswari Arwidarasti", "Masayuki Asahara", "Katla Ásgeirsdóttir", "Deniz Baran Aslan", "Cengiz Asmazoğlu", "Luma Ateyah", "Furkan Atmaca", "Mohammed Attia", "Aitziber Atutxa", "Liesbeth Augustinus", "Elena Badmaeva", "Keerthana Balasubramani", "Miguel Ballesteros", "Esha Banerjee", "Sebastian Bank", "Verginica Barbu Mititelu", "Starkaður Barkarson", "Rodolfo Basile", "Victoria Basmov", "Colin Batchelor", "John Bauer", "Seyyit Talha Bedir", "Juan Belieni", "Kepa Bengoetxea", "Yifat Ben Moshe", "Gözde Berk", "Yevgeni Berzak", "Irshad Ahmad Bhat", "Riyaz Ahmad Bhat", "Erica Biagetti", "Eckhard Bick", "Agnė Bielinskienė", "Kristín Bjarnadóttir", "Rogier Blokland", "Victoria Bobicev", "Loïc Boizou", "Emanuel Borges Völker", "Carl Börstell", "Cristina Bosco", "Gosse Bouma", "Sam Bowman", "Adriane Boyd", "Anouck Braggaar", "Kristina Brokaitė", "Aljoscha Burchardt", "Marie Candito", "Bernard Caron", "Gauthier Caron", "Lauren Cassidy", "Maria Clara Castro", "Tatiana Cavalcanti", "Gülşen Cebiroğlu Eryiğit", "Flavio Massimiliano Cecchini", "Giuseppe G. A. Celano", "Slavomír Čéplö", "Neslihan Cesur", "Savas Cetin", "Özlem Çetinoğlu", "Fabricio Chalub", "Liyanage Chamila", "Shweta Chauhan", "Ethan Chi", "Taishi Chika", "Yongseok Cho", "Jinho Choi", "Jayeol Chun", "Juyeon Chung", "Alessandra T. Cignarella", "Silvie Cinková", "Aurélie Collomb", "Çağrı Çöltekin", "Miriam Connor", "Daniela Corbetta", "Marine Courtin", "Mihaela Cristescu", "Philemon Daniel", "Elizabeth Davidson", "Leonel Figueiredo de Alencar", "Mathieu Dehouck", "Martina de Laurentiis", "Marie-Catherine de Marneffe", "Valeria de Paiva", "Mehmet Oguz Derin", "Elvis de Souza", "Arantza Diaz de Ilarraza", "Carly Dickerson", "Arawinda Dinakaramani", "Elisa Di Nuovo", "Bamba Dione", "Peter Dirix", "Kaja Dobrovoljc", "Timothy Dozat", "Kira Droganova", "Puneet Dwivedi", "Christian Ebert", "Hanne Eckhoff", "Sandra Eiche", "Marhaba Eli", "Ali Elkahky", "Binyam Ephrem", "Olga Erina", "Tomaž Erjavec", "Aline Etienne", "Wograine Evelyn", "Sidney Facundes", "Richárd Farkas", "Federica Favero", "Jannatul Ferdaousi", "Marília Fernanda", "Hector Fernandez Alcalde", "Jennifer Foster", "Cláudia Freitas", "Kazunori Fujita", "Katarína Gajdošová", "Daniel Galbraith", "Federica Gamba", "Marcos Garcia", "Moa Gärdenfors", "Sebastian Garza", "Fabrício Ferraz Gerardi", "Kim Gerdes", "Filip Ginter", "Gustavo Godoy", "Iakes Goenaga", "Koldo Gojenola", "Memduh Gökırmak", "Yoav Goldberg", "Xavier Gómez Guinovart", "Berta González Saavedra", "Bernadeta Griciūtė", "Matias Grioni", "Loïc Grobol", "Normunds Grūzītis", "Bruno Guillaume", "Céline Guillot-Barbance", "Tunga Güngör", "Nizar Habash", "Hinrik Hafsteinsson", "Jan Hajič", "Jan Hajič, jr.", "Mika Hämäläinen", "Linh Hà Mỹ", "Na-Rae Han", "Muhammad Yudistira Hanifmuti", "Takahiro Harada", "Sam Hardwick", "Kim Harris", "Dag Haug", "Johannes Heinecke", "Oliver Hellwig", "Felix Hennig", "Barbora Hladká", "Jaroslava Hlaváčová", "Florinel Hociung", "Petter Hohle", "Marivel Huerta Mendez", "Jena Hwang", "Takumi Ikeda", "Anton Karl Ingason", "Radu Ion", "Elena Irimia", "Ọlájídé Ishola", "Artan Islamaj", "Kaoru Ito", "Siratun Jannat", "Tomáš Jelínek", "Apoorva Jha", "Katharine Jiang", "Anders Johannsen", "Hildur Jónsdóttir", "Fredrik Jørgensen", "Markus Juutinen", "Hüner Kaşıkara", "Andre Kåsen", "Nadezhda Kabaeva", "Sylvain Kahane", "Hiroshi Kanayama", "Jenna Kanerva", "Neslihan Kara", "Ritván Karahóǧa", "Boris Katz", "Tolga Kayadelen", "Sarveswaran Kengatharaiyer", "Jessica Kenney", "Václava Kettnerová", "Jesse Kirchner", "Elena Klementieva", "Elena Klyachko", "Arne Köhn", "Abdullatif Köksal", "Kamil Kopacewicz", "Timo Korkiakangas", "Mehmet Köse", "Alexey Koshevoy", "Natalia Kotsyba", "Jolanta Kovalevskaitė", "Simon Krek", "Parameswari Krishnamurthy", "Sandra Kübler", "Adrian Kuqi", "Oğuzhan Kuyrukçu", "Aslı Kuzgun", "Sookyoung Kwak", "Veronika Laippala", "Lucia Lam", "Lorenzo Lambertino", "Tatiana Lando", "Septina Dian Larasati", "Alexei Lavrentiev", "John Lee", "Phương Lê Hồng", "Alessandro Lenci", "Saran Lertpradit", "Herman Leung", "Maria Levina", "Cheuk Ying Li", "Josie Li", "Keying Li", "Yixuan Li", "Yuan Li", "KyungTae Lim", "Bruna Lima Padovani", "Krister Lindén", "Nikola Ljubešić", "Olga Loginova", "Stefano Lusito", "Andry Luthfi", "Mikko Luukko", "Olga Lyashevskaya", "Teresa Lynn", "Vivien Macketanz", "Menel Mahamdi", "Jean Maillard", "Ilya Makarchuk", "Aibek Makazhanov", "Michael Mandl", "Christopher Manning", "Ruli Manurung", "Büşra Marşan", "Cătălina Mărănduc", "David Mareček", "Katrin Marheinecke", "Stella Markantonatou", "Héctor Martínez Alonso", "Lorena Martín Rodríguez", "André Martins", "Jan Mašek", "Hiroshi Matsuda", "Yuji Matsumoto", "Alessandro Mazzei", "Ryan McDonald", "Sarah McGuinness", "Gustavo Mendonça", "Tatiana Merzhevich", "Niko Miekka", "Karina Mischenkova", "Margarita Misirpashayeva", "Anna Missilä", "Cătălin Mititelu", "Maria Mitrofan", "Yusuke Miyao", "AmirHossein Mojiri Foroushani", "Judit Molnár", "Amirsaeid Moloodi", "Simonetta Montemagni", "Amir More", "Laura Moreno Romero", "Giovanni Moretti", "Keiko Sophie Mori", "Shinsuke Mori", "Tomohiko Morioka", "Shigeki Moro", "Bjartur Mortensen", "Bohdan Moskalevskyi", "Kadri Muischnek", "Robert Munro", "Yugo Murawaki", "Kaili Müürisep", "Pinkey Nainwani", "Mariam Nakhlé", "Juan Ignacio Navarro Horñiacek", "Anna Nedoluzhko", "Gunta Nešpore-Bērzkalne", "Manuela Nevaci", "Lương Nguyễn Thị", "Huyền Nguyễn Thị Minh", "Yoshihiro Nikaido", "Vitaly Nikolaev", "Rattima Nitisaroj", "Alireza Nourian", "Hanna Nurmi", "Stina Ojala", "Atul Kr. Ojha", "Hulda Óladóttir", "Adédayọ̀ Olúòkun", "Mai Omura", "Emeka Onwuegbuzia", "Noam Ordan", "Petya Osenova", "Robert Östling", "Lilja Øvrelid", "Şaziye Betül Özateş", "Merve Özçelik", "Arzucan Özgür", "Balkız Öztürk Başaran", "Teresa Paccosi", "Alessio Palmero Aprosio", "Anastasia Panova", "Hyunji Hayley Park", "Niko Partanen", "Elena Pascual", "Marco Passarotti", "Agnieszka Patejuk", "Guilherme Paulino-Passos", "Giulia Pedonese", "Angelika Peljak-Łapińska", "Siyao Peng", "Cenel-Augusto Perez", "Natalia Perkova", "Guy Perrier", "Slav Petrov", "Daria Petrova", "Andrea Peverelli", "Jason Phelan", "Jussi Piitulainen", "Rodrigo Pintucci", "Tommi A Pirinen", "Emily Pitler", "Magdalena Plamada", "Barbara Plank", "Thierry Poibeau", "Larisa Ponomareva", "Martin Popel", "Lauma Pretkalniņa", "Sophie Prévost", "Prokopis Prokopidis", "Adam Przepiórkowski", "Robert Pugh", "Tiina Puolakainen", "Sampo Pyysalo", "Peng Qi", "Andriela Rääbis", "Alexandre Rademaker", "Mizanur Rahoman", "Taraka Rama", "Loganathan Ramasamy", "Carlos Ramisch", "Fam Rashel", "Mohammad Sadegh Rasooli", "Vinit Ravishankar", "Livy Real", "Petru Rebeja", "Siva Reddy", "Mathilde Regnault", "Georg Rehm", "Ivan Riabov", "Michael Rießler", "Erika Rimkutė", "Larissa Rinaldi", "Laura Rituma", "Putri Rizqiyah", "Luisa Rocha", "Eiríkur Rögnvaldsson", "Ivan Roksandic", "Mykhailo Romanenko", "Rudolf Rosa", "Valentin Roșca", "Davide Rovati", "Ben Rozonoyer", "Olga Rudina", "Jack Rueter", "Kristján Rúnarsson", "Shoval Sadde", "Pegah Safari", "Benoît Sagot", "Aleksi Sahala", "Shadi Saleh", "Alessio Salomoni", "Tanja Samardžić", "Stephanie Samson", "Manuela Sanguinetti", "Ezgi Sanıyar", "Dage Särg", "Marta Sartor", "Mitsuya Sasaki", "Baiba Saulīte", "Yanin Sawanakunanon", "Shefali Saxena", "Kevin Scannell", "Salvatore Scarlata", "Nathan Schneider", "Sebastian Schuster", "Lane Schwartz", "Djamé Seddah", "Wolfgang Seeker", "Mojgan Seraji", "Syeda Shahzadi", "Mo Shen", "Atsuko Shimada", "Hiroyuki Shirasu", "Yana Shishkina", "Muh Shohibussirri", "Maria Shvedova", "Janine Siewert", "Einar Freyr Sigurðsson", "João Ricardo Silva", "Aline Silveira", "Natalia Silveira", "Maria Simi", "Radu Simionescu", "Katalin Simkó", "Mária Šimková", "Haukur Barri Símonarson", "Kiril Simov", "Dmitri Sitchinava", "Maria Skachedubova", "Aaron Smith", "Isabela Soares-Bastos", "Barbara Sonnenhauser", "Shafi Sourov", "Carolyn Spadine", "Rachele Sprugnoli", "Vivian Stamou", "Steinþór Steingrímsson", "Antonio Stella", "Abishek Stephen", "Milan Straka", "Emmett Strickland", "Jana Strnadová", "Alane Suhr", "Yogi Lesmana Sulestio", "Umut Sulubacak", "Shingo Suzuki", "Daniel Swanson", "Zsolt Szántó", "Chihiro Taguchi", "Dima Taji", "Yuta Takahashi", "Fabio Tamburini", "Mary Ann C. Tan", "Takaaki Tanaka", "Dipta Tanaya", "Mirko Tavoni", "Samson Tella", "Isabelle Tellier", "Marinella Testori", "Guillaume Thomas", "Sara Tonelli", "Liisi Torga", "Marsida Toska", "Trond Trosterud", "Anna Trukhina", "Reut Tsarfaty", "Utku Türk", "Francis Tyers", "Sveinbjörn Þórðarson", "Vilhjálmur Þorsteinsson", "Sumire Uematsu", "Roman Untilov", "Zdeňka Urešová", "Larraitz Uria", "Hans Uszkoreit", "Andrius Utka", "Elena Vagnoni", "Sowmya Vajjala", "Rob van der Goot", "Martine Vanhove", "Daniel van Niekerk", "Gertjan van Noord", "Viktor Varga", "Uliana Vedenina", "Giulia Venturi", "Eric Villemonte de la Clergerie", "Veronika Vincze", "Natalia Vlasova", "Aya Wakasa", "Joel C. Wallenberg", "Lars Wallin", "Abigail Walsh", "Jing Xian Wang", "Jonathan North Washington", "Maximilan Wendt", "Paul Widmer", "Shira Wigderson", "Sri Hartati Wijono", "Vanessa Berwanger Wille", "Seyi Williams", "Mats Wirén", "Christian Wittern", "Tsegay Woldemariam", "Tak-sum Wong", "Alina Wróblewska", "Mary Yako", "Kayo Yamashita", "Naoki Yamazaki", "Chunxiao Yan", "Koichi Yasuoka", "Marat M. Yavrumyan", "Arife Betül Yenice", "Olcay Taner Yıldız", "Zhuoran Yu", "Arlisa Yuliawati", "Zdeněk Žabokrtský", "Shorouq Zahra", "Amir Zeldes", "He Zhou", "Hanzhi Zhu", "Anna Zhuravleva", "Rayan Ziane" ]
https://www.semanticscholar.org/paper/1e105a47662affe083eaf110e5590ad4b9f7894d/
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
Preamble 1.0
Preambule 1.0
Preamble 1.0 is a multilingual annotated corpus of the preamble of the EU REGULATION 2020/2092 OF THE EUROPEAN PARLIAMENT AND OF THE COUNCIL with a manual annotation of sentence subjects. The corpus consists of four language versions of the preamble (Czech, English, French, Polish) and contains in total over 10 thousand words and over five hundred annotated subjects.
Preambule 1.0 je vícejazyčný anotovaný korpus preambule NAŘÍZENÍ EVROPSKÉHO PARLAMENTU A RADY 2020/2092 s ruční anotací větných subjektů. Korpus se skládá ze čtyř jazykových verzí preambule (české, anglické, francouzské, polské) a obsahuje celkem přes 10 tisíc slov a přes pět set anotovaných subjektů.
[ "Barbora Hladká", "Jiří Mírovský" ]
https://www.semanticscholar.org/paper/c83a6362881a7d042abd06690e2908f329494c41/
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
Annotating Attribution in Czech News Server Articles
Anotace přisuzování v článcích českého zpravodajského serveru
This paper focuses on detection of sources in the Czech articles published on a news server of Czech public radio. In particular, we search for attribution in sentences and we recognize attributed sources and their sentence context (signals). We organized a crowdsourcing annotation task that resulted in a data set of 2,167 stories with manually recognized signals and sources. In addition, the sources were classified into the classes of named and unnamed sources.
Tato práce se zaměřuje na detekci zdrojů v českých článcích publikovaných na zpravodajském serveru Českého veřejnoprávního rozhlasu. Hledáme zejména přiřazení ve větách a rozeznáváme přiřazené zdroje a jejich větný kontext(signály). Zorganizovali jsme crowdsourcingovou anotační úlohu, jejímž výsledkem byl datový soubor 2 167 článků s ručně rozpoznanými signály a zdroji. Zdroje byly navíc zařazeny do kategorií jmenovaných a nejmenovaných zdrojů.
[ "Barbora Hladká", "Jiří Mírovský", "Matyáš Kopp", "Václav Moravec" ]
https://www.semanticscholar.org/paper/f00f1fd2fdb620bd8c97a8167cef40ee198aae1d/
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
Quality and Efficiency of Manual Annotation: Data from the Pre-annotation Bias Experiment (part of the PDT-C 2.0 project)
Kvalita a efektivita manuální anotace: Data z experimentu zabývajícího se vlivem předanotace (součást projektu PDT-C 2.0)
Input data, individual experimental annotations, and a complete and detailed overview of the measured results related to the experiment dealing with the influence of automatic pre-annotation on the quality and efficiency of manual annotation efforts.
Vstupní data, jednotlivé experimentální anotace a úplný a podrobný přehled naměřených výsledků souvisejících s experimentem zabývajícím se vlivem automatické před-anotace na kvalitu a efektivitu anotačního úsilí.
[ "Marie Mikulová", "Milan Straka", "Jan Štěpánek", "Barbora Štěpánková", "Jan Hajič" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
Quality and Efficiency of Manual Annotation: Pre-annotation Bias
Kvalita a efektivita ruční anotace: zkreslení před-anotací
This paper presents an analysis of annotation using an automatic pre-annotation for a mid-level annotation complexity task - dependency syntax annotation. It compares the annotation efforts made by annotators using a pre-annotated version (with a high-accuracy parser) and those made by fully manual annotation. The aim of the experiment is to judge the final annotation quality when pre-annotation is used. In addition, it evaluates the effect of automatic linguistically-based (rule-formulated) checks and another annotation on the same data available to the annotators, and their influence on annotation quality and efficiency. The experiment confirmed that the pre-annotation is an efficient tool for faster manual syntactic annotation which increases the consistency of the resulting annotation without reducing its quality.
Tento článek představuje analýzu anotace pomocí automatické před-anotace pro úkol anotace závislostní syntaxe. Porovnává úsilí anotátorů, kteří používají před-anotovanou verzi (s vysoce přesným parserem), a úsilí vytvořené plně ruční anotací. Cílem experimentu je posoudit výslednou kvalitu anotace při použití před-anotace. Kromě toho hodnotí vliv automatických lingvisticky založených (pravidlově formulovaných) kontrol a jiné anotace na stejných datech, kterou mají anotátoři k dispozici, a jejich vliv na kvalitu a efektivitu anotace. Experiment potvrdil, že pre-anotace je účinným nástrojem pro rychlejší manuální syntaktickou anotaci, která zvyšuje konzistenci výsledné anotace bez snížení její kvality.
[ "Marie Mikulová", "Milan Straka", "Jan Štěpánek", "Barbora Štěpánková", "Jan Hajič" ]
https://www.semanticscholar.org/paper/f2942feb0ff284e867175894191a9a93f7ac0591/
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
Towards Reciprocity in Logic and in Linguistics
K pojetí reciprocity v logice a v lingvistice
Reciprocity / mutuality / symmetry: intensively studied in linguistics and in formal semantics (among other fields). However, there are still open questions concerning esp. terminological aspects and a proper test allowing us to reliably distinguish predicates and constructions with reciprocal meaning. According to Gleitman & Partee (2020), "… considerations suggest that the standard logical definitions of symmetrical and its relatives are by themselves too limited for linguists’ needs. Common usage offers clues to further notions that are descriptively and perhaps explanatorily valuable."
Koncepty reciprocity/vzájemnosti/symetrie jsou intenzivně studovány v lingvistice a ve formální sémantice (mimo jiné obory). Stále však existují otevřené otázky týkající se zejména terminologických aspektů a řádného testu, který by umožňoval spolehlivě rozlišovat predikáty a konstrukce s recipročním významem. Podle Gleitman & Partee (2020) standardní logické definice symetrie a příbuzných konceptů jsou samy o sobě příliš omezující pro potřeby lingvistů. Jazyková data nabízejí vodítka k dalšímu vymezení vhodných pojmům, které umožní adekvátní zachycení recipročních konstrukcí.
[ "Markéta Lopatková", "Václava Kettnerová" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
cs
2022
VALLEX 4.5
VALLEX 4.5
The Valency Lexicon of Czech Verbs is a collection of linguistically annotated data and documentation; it provides a formal, machine-readable description of valency frames of Czech verbs and additional syntactico-semantic information useful for the analysis and synthesis of Czech texts as well as other applied tasks in NLP. It covers the common senses of the most frequent Czech verbs (in total over 11,080 senses of almost 4,700 lemmas, i.e., more than 6,850 verb senses counting perfective and imperfective verbs as forming a single lexeme). VALLEX 4.5 is an enhanced successor of VALLEX 4.0, 3.5, and 3.0. In addition to the information stored there, VALLEX 4.5 provides a detailed description of reflexive verbs, i.e., verbs with the reflexive se or si as an obligatory part of their verb lexemes. VALLEX 4.5 covers 1,525 reflexive verbs in 1,545 lexical units (2,501 when aspectual counterparts counted separately). VALLEX 4.5 provides also enhanced characteristics of verbs expressing reciprocity and reflexivity.
Valenční slovník českých sloves (též VALLEX 4.5) je souborem lingvisticky anotovaných dat a dokumentace. Slovník poskytuje informace o valenční struktuře českých sloves v jejich jednotlivých významech, které charakterizuje pomocí glos a příkladů; tyto základní údaje jsou doplněny o některé další syntaktické a sémantické charakteristiky. Tuto informace jsou dostupné pro čtenáře, mají však i formální, strojově čitelnou podobu vhodnou pro aplikované úlohy v NLP. VALLEX zachycuje nejčastější česká slovesa sloves (celkem přes 11 080 významů pro téměř 4 700 českých sloves, což představuje více než 6 850 významů, sdruží-li se vidové protějšky sloves do lexémů). VALLEX 4.5 je rozšířeným nástupcem VALLEXu 4.0, 3.5 a 3.0. - VALLEX 4.5 navíc přináší charakteristiku reflexivních sloves, tj. sloves s reflexivním se nebo si jako povinnou součást slovesných lexémů. VALLEX 4.5 zahrnuje 1 525 reflexivních sloves v 1 545 lexikálních jednotkách (2 501 sloves, jsou-li vidové protějšky počítány samostatně), nově též zpracovává i slovesa vyjadřující reflexivní a reciproční významy.
[ "Markéta Lopatková", "Václava Kettnerová", "Jiří Mírovský", "Anna Vernerová", "Eduard Bejček", "Zdeněk Žabokrtský" ]
https://www.semanticscholar.org/paper/f7ec8415b1e51d3dd8df89cd4b230077937a0f8e/
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
Global Variants in the Czech Language
Globální varianty v českém jazyce
There are words written in several different ways in Czech, e.g., lampion ∼ lampión (lampion). This variability may occur in either some inflectional word- forms (inflectional variants), cf. hradu ∼ hradě in the locative case of the noun hrad (castle), or across the inflectional wordforms and derivatives (global variants), cf. fantazijní ∼ fantasijní in the adjective derived from the noun fantazie ∼ fantasie (fantasy). It is reasonable to distinguish the global variants as different words but to have formal means that interconnect them in the Natural Language Processing systems and resources. In this paper, we describe the identification of global variants in the Czech vocabulary and summarise new changes in the MorfFlex CZ dictionary and DeriNet lexicon concerning this type of variants. We reviewed several typical patterns within global variants captured in the available resources and combined a set of regular expressions with manual annotations to achieve the highest precision of the identification.
V češtině se vyskytují slova psaná několika různými způsoby, např. lampion ∼ lampión. Tato variabilita se může vyskytovat buď v některých flektivních slovních formách (flektivní varianty), srov. hradu ∼ hradě v lokálu podstatného jména hrad, nebo napříč flektivními slovními formami a odvozeninami (globální varianty), srov. fantazijní ∼ fantasijní v přídavném jméně odvozeném od podstatného jména fantazie ∼ fantasie. Je rozumné rozlišovat globální varianty jako různá slova, ale mít formální prostředky, které je propojují v systémech a zdrojích zpracování přirozeného jazyka. V tomto příspěvku popisujeme identifikaci globálních variant v českém slovníku a shrnujeme nové změny ve slovníku MorfFlex CZ a lexikonu DeriNet týkající se tohoto typu variant. Analyzovali jsme několik typických vzorů v rámci globálních variant zachycených v dostupných zdrojích a zkombinovali sadu regulárních výrazů s manuálními anotacemi, abychom dosáhli co nejvyšší přesnosti identifikace.
[ "Jaroslava Hlaváčová", "Lukáš Kyjánek", "Magda Ševčíková" ]
https://www.semanticscholar.org/paper/f06df4118b9d85b1e8a27641e8b5f310ac6a6eb2/
null
null
null
null
null
null
null
null
null
null
null
null
cs
2022
Readability as a scientific concept in psychology, linguistics and mathematics
Jak se na srozumitelnost dívá věda? Srozumitelnost v zrcadle psychologie, lingvistiky a matematiky
Lecture for civil servants and lawyers at the Round Table of the Office of the Public Defender of Rights. It aimed at presenting the research of readability and reading comprehension to an audience with a different background.
Přednáška pro státní úředníky a právníky u Kulatého stolu Kanceláře veřejného ochránce práv. Jejím účelem bylo publikum z jiných oborů seznámit s výzkumem srozumitelnosti textů a procesu čtení a porozumění z hlediska lingvistiky i kognitivní vědy.
[ "Silvie Cinková" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
cs
2022
Digital Humanities courses in Czech Tertiary Education Described by Latent Semantic Analysis
Výuka digitálních humanitních věd na českých veřejných vysokých školách podle latentní sémantické analýzy
Our study was created as the first attempt to map the Czech pedagogy of Digital Humanities in 2018. We believe that Czech memory institutions are already well integrated into European infrastructure projects (eg Library of the Academy of Sciences, National Library and National Museum in Europeana) and Czech researchers have also successfully joined the international DH community (as evidenced by their contributions in this volume). So far, however, we know little about the DH paradigm in the teaching of humanities at Czech universities. To aggregate the information, we used Latent semantic analysis on the names of the courses and institutions where they are taught.
Naše studie vznikla jako první pokus zmapovat podhoubí české pedagogiky digitálních humanitních věd v roce 2018. Máme za to, že české paměťové instituce jsou už dobře integrovány do evropských infrastrukturních projektů (například Knihovna Akademie věd ČR, Národní knihovna a Národní muzeum v projektu Europeana ) a čeští badatelé se již také úspěšně připojili k mezinárodní DH komunitě (o čemž svědčí také jejich příspěvky v tomto svazku). Zatím však málo víme o DH paradigmatu ve výuce humanitních oborů na českých vysokých školách.Proto jsme během roku 2018 ručně prohledali aktuální studijní katalogy většiny českých veřejných vysokých škol s cílem shromáždit kontakty na vyučující, kteří by mohli stát u zrodu digitálně humanitní pedagogické komunity v Česku, a udělat si představu o jejich specializacích. K agregaci informací jsme použili Latentně sémantickou analýzu na názvech kurzů a institucí, kde se vyučují.
[ "Silvie Cinková", "Jan Škvrňák", "Michael Škvrňák" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
CLS Infra Training School on Data and Annotation
CLS Infra Letní škola o datech a anotaci
Keen to get a flying start in Digital Humanities? You can build, edit, annotate, and query a text corpus without a single line of code! Structure texts with the XML-TEI. Run an NLP tool to add linguistic information. Tackle real research questions concerning Shakespeare’s dramas or your own data by mastering CQL and Universal Dependencies.
Rádi byste začali v Digital Humanities? Můžete vytvářet, upravovat, komentovat a dotazovat se na textový korpus bez jediného řádku kódu! Strukturujte texty pomocí XML-TEI. Spusťte nástroj NLP pro přidání jazykových informací. Vyřešte skutečné výzkumné otázky týkající se Shakespearových dramat nebo svých vlastních dat zvládnutím CQL a Universal Dependencies.
[ "Silvie Cinková", "Maarten Janssen", "Michal Křen", "Václav Cvrček" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
Action meanings in noun/verb conversion: Native and foreign word-formation in Czech
Dějové významy v konverzi mezi substantivem a slovesem: Tvoření slov s domácími a přejatými kořeny v češtině
The study deals with Czech conversion pairs of a noun and a verb, both of which denote actions (test ‘test.n’ – testovat ‘to test’). Elaborating on previous research on prototypical verb-to-noun and noun-to-verb conversion in Czech, the direction in these pairs is determined based on whether the verb forms its aspectual counterpart by changing the theme (which is characteristic of the deverbal direction), or whether the suffixed counterpart is not available (typical of denominal verbs). The analysis, carried out on a corpus sample of 1,300 action nouns and directly related verbs, demonstrates that pairs with native roots mostly conform to the deverbal pattern, whereas the denominal direction applies to a smaller subset of the native sample but clearly prevails in the data with foreign roots. The denominal direction ascribed to foreign pairs is consistent with the typological hypothesis that verbs are borrowed rather as nouns and subsequently turned into verbs in the target language.
Studie se zabývá českými konverzními dvojicemi, v nichž podstatné jméno i sloveso mají dějový význam (test 'zkoušet.n' - testovat 'zkoušet'). V návaznosti na předchozí výzkum prototypických případů deverbální a denominální konverze v češtině je směr v těchto dvojicích určován na základě toho, zda sloveso tvoří svůj vidový protějšek změnou kmenotvorné přípony (což je charakteristické pro deverbální směr), nebo zda sufixální protějšek není k dispozici (typické pro denominální slovesa). Analýza provedená na korpusovém vzorku 1 300 podstatných jmen s dějovým významem a přímo souvisejících sloves ukazuje, že dvojice založené na domácích kořenech mají většinou rysy deverbálního tvoření, zatímco denominální směr se uplatňuje v menší skupině domácích dvojic a v datech s cizími kořeny jasně převažuje. Denominální směr přisuzovaný dvojicím s cizími kořeny je v souladu s typologickou hypotézou, že slovesa jsou přejímána do češtiny spíše jako podstatná jména a následně se v cílovém jazyce mění na slovesa.
[ "Magda Ševčíková" ]
https://www.semanticscholar.org/paper/f8aadd271059b2d4269ce56249d2d43c38a97102/
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
Annotation of semantic categories in noun/verb conversion: A pilot experiment in Czech
Anotace sémantických kategorií v konverzním tvoření podstatných jmen a sloves: Pilotní experiment na češtině
The paper presents and evaluates an experiment in annotating semantic categories expressed by conversion in Czech. Based on existing repertoires of semantic labels applied to conversion, the design, implementation, and results of the annotation experiment are described, and discussed in the context of previous, exclusively qualitative descriptions of conversion in Czech. After eliminating pairs in which the semantic relationship between noun and verb was not perceived in synchrony, the resulting set of 2,000 pairs contained only those with a link between at least one noun sense and one verb sense. For each noun/verb pair, 100 corpus concordances were assigned a semantic value by two human annotators in parallel (native speakers of Czech, linguists). The inter-annotator agreement is calculated using Krippendorff’s alpha coefficient, which reduces the effect of chance agreement. The results range from moderate to substantial agreement depending on the semantic group.
Příspěvek představuje a vyhodnocuje paralelní anotaci významů, které jsou v češtině vyjadřovány při tvoření konverzí.
[ "Magda Ševčíková" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
Conversion of action nouns to verbs: Insights from Czech
Tvoření sloves z dějových substnativ konverzí: Případ češtiny
Although actions are prototypical verbal concepts, there are verbs that are assumed to be derived or converted from nouns with action meanings, cf. apology > to apologize or attack.n > to attack. Conversion of action nouns to verbs, discussed as the PERFORMATIVE category by Plag (1999) or Bauer et al. (2013), has been attested since the earliest period documented by the Oxford English Dictionary and, remarkably, have become predominant among verbs converted from nouns in English during the 20th century (Gottfurcht 2008). The present paper deals with Czech conversion pairs, where both the noun and the verb denote an action (e.g. řez ‘cut.n’ – řezat ‘to cut’, test ‘test.n’ – testovat ‘to test’). Elaborating on our previous research on word formation without derivational affixes in Czech, phonological and morphological features of nouns and verbs are employed to determine the direction of conversion in individual pairs.
Přestože dějový význam je prototypicky přisuzován slovesu, existují slovesa, u nichž se předpokládá, že jsou odvozena nebo konvertována od podstatných jmen s dějovým významem. Předkládaný příspěvek se zabývá českými konverzními dvojicemi, kde jak podstatné jméno, tak sloveso vyjadřují děj (např. řez - řezat, test - testovat). V návaznosti na náš předchozí výzkum tvoření slov bez derivačních afixů v češtině jsou pro určení směru konverze v jednotlivých dvojicích využity fonologické a morfologické vlastnosti podstatných jmen a sloves.
[ "Magda Ševčíková" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
Paradigms in English and Czech noun/verb conversion: A contrastive study of lexemes with borrowed roots
Paradigmata v konverzním tvoření substantiv a sloves v angličtině a češtině: Kontrastivní studie lexémů s přejatými kořeny
The study deals with English noun/verb conversion pairs that have both formally and semantically close counterpart pairs in Czech. The study’s aim is to examine how these nouns and verbs, linked with similar semantic relations in English and Czech, are accommodated in the two languages with different morphological structures and conversion playing a different role. The noun/verb pairs, extracted from the British National Corpus and from the SYN2000 corpus, are analysed as two-cell paradigms and examined along with selected derivatives. The data suggest that in the Czech sample, nominals are preferred over verbs in expressing the particular meanings and most verbs appear as denominal formations, often differently from their English counterparts.
Studie se zabývá anglickými konverzními dvojicemi podstatných a sloves, které mají formálně i významově blízké protějšky v češtině. Cílem studie je prozkoumat, jak se tato substantiva a slovesa, která jsou v angličtině a češtině spojena podobnými sémantickými vztahy, chovají v jazycích, které mají odlišnou morfologickou strukturu a v jejichž slovotvorných systémech hraje konverze odlišnou roli. Dvojice substantiv a sloves, extrahované z Britského národního korpusu a z korpusu SYN2000, jsou popisována jako dvoučlenná paradigmata a zkoumány spolu s vybranými deriváty. Data ukazují, že v českých dvojicích jsou pro vyjadřování konkrétních významů preferována substantiva před slovesy a většina sloves je používána jako denominální formace, často odlišně od anglických protějšků.
[ "Magda Ševčíková", "Hana Hledíková" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
TEI and Git in ParlaMint: Collaborative Development of Language Resources
TEI a Git v ParlaMint: Spolupráce při rozvoji jazykových zdrojů
This paper discusses the encoding, validation and development of language resources of the completed ParlaMint I and on-going ParlaMint II CLARIN projects, which centre on the collaborative development of TEI-encoded corpora of parliamentary proceedings. We introduce the use of TEI ODD to write the encoding guidelines and formal XML schemas for validation. We motivate and explain using Git to develop of encoding schemas and language resources. Apart from revision control, issues and publishing documentation, we also emphasise GitHub actions with their ability to integrate program code execution into the data submission process. The paper is written with a view to introducing SSH scientists to the two environments, as they can be valuable items in the toolbox for compiling language resources, especially in a collaborative setting.
Tento článek pojednává o kódování, validaci a vývoji jazykových zdrojů dokončeného projektu ParlaMint I a probíhajícího projektu ParlaMint II CLARIN, které se zaměřují na společný vývoj korpusů parlamentních jednání kódovaných TEI. Zavádíme využití TEI ODD pro psaní pokynů pro kódování a formálních XML schémat pro validaci. Motivujeme a vysvětlujeme použití Gitu pro vývoj kódovacích schémat a jazykových zdrojů. Kromě kontroly revizí, problémů a publikační dokumentace klademe důraz také na akce GitHubu s jejich schopností integrovat provádění programového kódu do procesu předkládání dat. Článek je napsán s cílem seznámit vědce SSH s oběma prostředími, protože mohou být cennými položkami v sadě nástrojů pro kompilaci jazykových zdrojů, zejména v kolaborativním prostředí.
[ "Tomaž Erjavec", "Matyáš Kopp" ]
https://www.semanticscholar.org/paper/b97d2f2549afca793759ee9f12b2a83b028eace0/
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
Language Technologies for the Creation of Multilingual Terminologies. Lessons Learned from the SSHOC Project
Jazykové technologie pro tvorbu vícejazyčných terminologií. Poznatky z projektu SSHOC
This paper is framed in the context of the SSHOC project and aims at exploring how Language Technologies can help in promoting and facilitating multilingualism in the Social Sciences and Humanities (SSH). Although most SSH researchers produce culturally and societally relevant work in their local languages, metadata and vocabularies used in the SSH domain to describe and index research data are currently mostly in English. We thus investigate Natural Language Processing and Machine Translation approaches in view of providing resources and tools to foster multilingual access and discovery to SSH content across different languages. As case studies, we create and deliver as freely, openly available data a set of multilingual metadata concepts and an automatically extracted multilingual Data Stewardship terminology. The two case studies allow as well to evaluate performances of state-of-the-art tools and to derive a set of recommendations as to how best apply them. Although not adapted to the specific domain, the employed tools prove to be a valid asset to translation tasks. Nonetheless, validation of results by domain experts proficient in the language is an unavoidable phase of the whole workflow.
Tento článek je zasazen do kontextu projektu SSHOC a jeho cílem je prozkoumat, jak mohou jazykové technologie pomoci při podpoře a usnadnění vícejazyčnosti v sociálních a humanitních vědách (SSH). Přestože většina výzkumných pracovníků v oblasti SSH vytváří kulturně a společensky relevantní práce ve svých místních jazycích, metadata a slovníky používané v oblasti SSH k popisu a indexování výzkumných dat jsou v současné době většinou v angličtině. Zkoumáme proto přístupy zpracování přirozeného jazyka a strojového překladu s cílem poskytnout zdroje a nástroje na podporu vícejazyčného přístupu k obsahu SSH a jeho vyhledávání v různých jazycích. Jako případové studie vytváříme a poskytujeme jako volně dostupná data soubor vícejazyčných metadatových konceptů a automaticky extrahovanou vícejazyčnou terminologii Data Stewardship. Tyto dvě případové studie umožňují také vyhodnotit výkonnost nejmodernějších nástrojů a odvodit soubor doporučení, jak je nejlépe použít. Přestože nejsou přizpůsobeny konkrétní doméně, použité nástroje se ukázaly být platným přínosem pro překladatelské úlohy. Nicméně ověření výsledků experty na danou doménu, kteří ovládají daný jazyk, je nevyhnutelnou fází celého pracovního postupu.
[ "Federica Gamba", "Francesca Frontini", "Daan Broeder", "Monica Monachini" ]
https://www.semanticscholar.org/paper/1b215e156efb6a792cf61d5a49204912a4a47ac5/
null
null
null
null
null
null
null
null
null
null
null
null
cs
2022
70/70 Rudolf Rosa
70/70: Rudolf Rosa
The faculty celebrates its anniversary, turning 70 this September. We bring another part of the poll, in which seven dozen personalities from scientists, teachers, students and graduates wish Matfyza a happy birthday.
Fakulta slaví kulaté výročí, letos v září jí bude už 70 let. Přinášíme další díl ankety, ve které přeje Matfyzu k narozeninám sedm desítek osobností z řad vědců, učitelů, studentů a absolventů.
[ "Rudolf Rosa" ]
https://www.semanticscholar.org/paper/29e478b93572b02399b4d14a1721f3a5da983075/
null
null
null
null
null
null
null
null
null
null
null
null
cs
2022
AI in theatre
AI v divadle
You can go to Švanda's theatre to see "AI: When a Robot Writes a Play". The actors are human, but 90% of the script was generated by our THEaiTRobot tool. How did he do it? And did he do it? You will learn how the language model GPT-2 works and how we modified it to generate plays. You will see a sample of the play. And you will get the opportunity to generate your own theatre script.
Do Švandova divadla můžete zajít na představení "AI: Když robot píše hru". Herci jsou lidští, ale 90% scénáře vygeneroval náš nástroj THEaiTRobot. Jak to udělal? A povedlo se mu to? Dozvíte se, jak funguje jazykový model GPT-2 a jak jsme ho upravili pro generování divadelních her. Uvidíte ukázku z divadelní hry. A dostanete příležitost si sami vygenerovat vlastní divadelní scénář.
[ "Rudolf Rosa" ]
https://www.semanticscholar.org/paper/b26fcebb85817856aa8ed430715118f0aadb7584/
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
Elitr LangTools and THEaiTRE script generation
Elitr LangTools a generování scénářů THEaiTRE
Presentation of Elitr language tools and THEaiTRE script generation for EUROSAI congress attendees.
Prezentace jazykových nástrojů systému Elitr a generování scénářů THEaiTRE pro účastníky EUROSAI kongresu.
[ "Rudolf Rosa" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
cs
2022
How GPT-2 generates plays
Jak GPT-2 generuje hry
Brief explanation of the process of creation of the automatically generated script of the play.
Stručné vysvětlení procesu vzniku automaticky generovaného scénáře divadelní hry.
[ "Rudolf Rosa" ]
https://www.semanticscholar.org/paper/ee5ba5bc52dd108402560c1a00aea8992df04a4c/
null
null
null
null
null
null
null
null
null
null
null
null
cs
2022
How Artificial Intelligence Wrote a Play
Jak umělá inteligence napsala divadelní hru
You can go to Švanda Theatre to see "AI: When a Robot Writes a Play". The actors are human, but 90% of the script was generated by our THEaiTRobot tool. How did it do it? And did it succeed? You will learn how the language model GPT-2 works and how we modified it for generating plays. You will see a sample of the play. And you will get the opportunity to generate your own theatre script.
Do Švandova divadla můžete zajít na představení "AI: Když robot píše hru". Herci jsou lidští, ale 90% scénáře vygeneroval náš nástroj THEaiTRobot. Jak to udělal? A povedlo se mu to? Dozvíte se, jak funguje jazykový model GPT-2 a jak jsme ho upravili pro generování divadelních her. Uvidíte ukázku z divadelní hry. A dostanete příležitost si sami vygenerovat vlastní divadelní scénář.
[ "Rudolf Rosa" ]
https://www.semanticscholar.org/paper/155108784e14747496af8ff5a1e2d40d72d2b7fb/
null
null
null
null
null
null
null
null
null
null
null
null
cs
2022
Presenting the exhibit Life with Artificial Intelligence
Představení exponátu Život s umělou inteligencí
AI, artificial intelligence, smart devices, machine learning... We hear these words around us more and more. Artificial intelligence is gradually infiltrating many fields of human activity, from the development of new drugs to the writing of plays to religion. But is it all the same artificial intelligence? What does it do? How is it used? And what exactly is AI?
AI, umělá inteligence, chytrá zařízení, strojové učení... Tato slova kolem sebe slýcháme stále častěji. Umělá inteligence postupně proniká do mnoha oborů lidské činnosti, od vývoje nových léků přes psaní divadelních her až po náboženství. Je to ale všechno stejná umělá inteligence? Co umí? Jak se používá? A co to vlastně je AI?
[ "Rudolf Rosa" ]
https://www.semanticscholar.org/paper/121482e57280b433a6a843deb7ec71e1ae5f88f9/
null
null
null
null
null
null
null
null
null
null
null
null
cs
2022
THEaiTRE project: automatic interactive generation of theatre play scripts
Projekt THEaiTRE: automatické interaktivní generování scénářů divadelních her
As part of the showroom of the conference Innovative Vision of Prague 2030, we present the introduction of the first Czech theatre play written by artificial intelligence.
V rámci showroomu konference Inovační vize Prahy 2030 představujeme uvedení první české divadelní hry napsané umělou inteligencí.
[ "Rudolf Rosa" ]
https://www.semanticscholar.org/paper/5cf787dcb21cc9db4a689c57702d0a247222a3ff/
null
null
null
null
null
null
null
null
null
null
null
null
en
2022
THEaiTRE
THEaiTRE
We inform about the ending THEaiTRE project, focusing on hierarchical generation, dramatic situations, and future projects.
Informujeme o končícím projektu THEaiTRE, se zaměřením na hierarchické generování, dramatické situace a budoucí projekty.
[ "Rudolf Rosa" ]
https://www.semanticscholar.org/paper/5cf787dcb21cc9db4a689c57702d0a247222a3ff/
null
null
null
null
null
null
null
null
null
null
null
null
cs
2022
THEaiTRE
THEaiTRE
We present our THEaiTRE project to the employees of the LMC company as a part of their internal workshop.
Představíme náš projekt THEaiTRE zaměstnancům společnosti LMC v rámci jejich interního workshopu.
[ "Rudolf Rosa" ]
https://www.semanticscholar.org/paper/5cf787dcb21cc9db4a689c57702d0a247222a3ff/
null
null
null
null
null
null
null
null
null
null
null
null