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
2021
Adaptation of Classic Readability Metrics to Czech
Adaptace klasických metrik srozumitelnosti textu na češtinu
We have fitted four classic readability metrics to Czech, using InterCorp (a parallel corpus with manual sentence alignment), CzEng 2.0 (a large parallel corpus of crawled web texts), and the optimize.curve fit algorithm from the SciPy library. The adapted metrics are: Flesch Reading Ease, Flesch-Kincaid Grade Level, Coleman-Liau Index, and Automated Readability Index. We describe the details of the procedure and present satisfactory results. Besides, we discuss the sensitivity of these metrics to text paraphrases and correlation of readability scores with empirically observed reading comprehension, as well as the adaptation of Flesch Reading Ease to Czech from Russian.
Adaptovali jsme na češtinu čtyři klasické metriky srozumitelnosti na základě dat InterCorp (paralelní korpus s ručním zarovnáním vět)a CzEng 2.0 (velký paralelní korpus procházených webových textů) a algoritmu optimize.curve z knihovny SciPy. Adaptované metriky jsou: Flesch Reading Ease, Flesch-Kincaid Grade Level, Coleman-Liau Index a Automated Readability Index. Popisujeme podrobnosti postupu a předkládáme uspokojivé výsledky. Kromě toho diskutujeme citlivost těchto metrik na textové parafráze a korelaci skóre srozumitelnosti s empiricky pozorovaným porozuměním čtenému a také adaptaci Flesch Reading Ease na češtinu z ruštiny.
[ "Klára Bendová", "Silvie Cinková" ]
https://www.semanticscholar.org/paper/222f53397da283d15a6276aa9a3063fa4d50fc33/
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
Accuracy in Neural Text Generation
Přesnost v neuronových generátorech textu
While neural-network-based approaches have significantly increased the fluency of machine-generated text, they are prone to accuracy errors, such as leaving out parts of the input (omissions) or generating output that is not grounded in any input (hallucination). This problem becomes especially apparent with lower amounts of training data, generation of longer texts, and noise in the training data. All of these conditions are very common for natural language generation (NLG) tasks. In this talk, I will show several example approaches aiming to produce more accurate outputs within an NLG system, or to detect errors in an NLG system output. I will specifically focus on sequence-to-sequence neural architectures with pretrained language models and their extensions. I will mostly discuss generation of text from structured data, but I will also make comparisons to the situation in summarization or dialogue response generation.
Přestože přístupy založené na neuronových sítích výrazně zvýšily plynulost strojově generovaného textu, jsou náchylné k chybám v přesnosti, jako je vynechávání částí vstupu (opomenutí) nebo generování výstupu, který nemá oporu v žádném vstupu (halucinace). Tento problém se projevuje zejména při menším množství trénovacích dat, generování delších textů nebo šumu v trénovacích datech. Všechny tyto podmínky jsou pro úlohy generování přirozeného jazyka (NLG) velmi časté. V této přednášce ukážu několik příkladů přístupů zaměřených na vytváření přesnějších výstupů v rámci systému NLG nebo na odhalování chyb ve výstupech systému NLG. Konkrétně se zaměřím na neuronové architektury sequence-to-sequence s předtrénovanými jazykovými modely a jejich rozšíření. Budu se zabývat především generováním textu ze strukturovaných dat, ale uvedu i srovnání se situací při sumarizaci nebo generování dialogových odpovědí.
[ "Ondřej Dušek" ]
https://www.semanticscholar.org/paper/998592c0f3aaa6d64a1a3a79aa9e9caa3c7a5633/
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
Better Supervision for End-to-end Neural Dialogue Systems
Lepší supervize pro end-to-end neuronové dialogové systémy
While end-to-end neural models have been the research trend in task-oriented dialogue systems in the past years, they still suffer from significant problems: The neural models often produce replies inconsistent with past dialogue context or database results, their replies may be dull and formulaic, and they require large amounts of annotated data to train. In this talk, I will present two of our recent experiments that aim at solving these problems. First, our end-to-end neural system AuGPT based on the GPT-2 pretrained language model aims at consistency and variability in dialogue responses by using massive data augmentation and filtering as well as specific auxiliary training objectives which check for dialogue consistency. It reached favorable results in terms of both automatic metrics and human judgments (in the DSTC9 competition). Second, we designed a system that is able to discover relevant dialogue slots (domain attributes) without any human annotation. It uses weak supervision from generic linguistic annotation models (semantic parser, named entities), which is further filtered and clustered. We train a neural slot tagger on the discovered slots, which then reaches state-of-the-art results in dialogue slot tagging without labeled training data. We further show that the discovered slots are helpful for training an end-to-end neural dialogue system.
Přestože neuronové modely typu end-to-end jsou v posledních letech trendem výzkumu v oblasti dialogových systémů orientovaných na úkoly, stále se potýkají s významnými problémy: Neuronové modely často vytvářejí odpovědi, které neodpovídají minulému kontextu dialogu nebo výsledkům databáze, jejich odpovědi mohou být nudné a formulaické a k jejich trénování je zapotřebí velké množství anotovaných dat. V této přednášce představím dva z našich nedávných experimentů, jejichž cílem je tyto problémy vyřešit. Zaprvé, náš end-to-end neuronový systém AuGPT založený na předtrénovaném jazykovém modelu GPT-2 se zaměřuje na konzistenci a variabilitu odpovědí v dialogu pomocí masivního rozšiřování a filtrování dat a také specifických pomocných trénovacích cílů, které kontrolují konzistenci dialogu. Dosáhl příznivých výsledků, pokud jde o automatické metriky i lidské hodnocení (v soutěži DSTC9). Za druhé jsme navrhli systém, který je schopen objevit relevantní dialogové sloty (atributy domény) bez jakékoli lidské anotace. Využívá slabou supervizi z obecných lingvistických anotačních modelů (sémantický parser, pojmenované entity), které dále filtruje a shlukuje. Na objevených slotech trénujeme neuronový tagger slotů, který pak dosahuje nejlepších výsledků v tagování dialogových slotů bez označených trénovacích dat. Dále ukazujeme, že objevené sloty jsou užitečné pro trénování neuronového dialogového systému end-to-end.
[ "Ondřej Dušek" ]
https://www.semanticscholar.org/paper/55289d3feef4bc1e4ff17008120e371eb7f55a24/
null
null
null
null
null
null
null
null
null
null
null
null
cs
2021
THEaiTRobot demo for the exhibition Who is afraid of artificial intelligence
THEaiTRobot demo pro výstavu Kdo se bojí umělé inteligence
Who's afraid of artificial intelligence? Come to the Goethe-Institut thematic space and find out how literature is made with the help of robots! https://www.goethe.de/ins/cz/en/ver.cfm?event_id=22345514
Kdo se bojí umělé inteligence. Přijďte do tematického prostoru Goethe-Institutu a zjistěte, jak s pomocí robotů vzniká literatura! https://www.goethe.de/ins/cz/cs/ver.cfm?event_id=22345514
[ "Alisa Zakhtarenko", "Rudolf Rosa", "Ondřej Dušek" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
FINDINGS OF THE IWSLT 2021 EVALUATION CAMPAIGN
Výsledky evaluační kampaně IWSLT 2021
The evaluation campaign of the International Conference on Spoken Language Translation (IWSLT 2021) featured this year four shared tasks: (i) Simultaneous speech translation, (ii) Offline speech translation, (iii) Multilingual speech translation, (iv) Low-resource speech translation. A total of 22 teams participated in at least one of the tasks. This paper de- scribes each shared task, data and evaluation metrics, and reports results of the received submissions.
Hodnotící kampaň Mezinárodní konference o překladu z mluveného jazyka (IWSLT 2021) letos představila čtyři společné úkoly: (i) Simultánní překlad mluvené řeči, (ii) Offline překlad mluvené řeči, (iii) Vícejazyčný překlad řeči, (iv) překlad řeči s malým množstvím zdrojů. Celkem se zúčastnilo 22 týmů alespoň na jedné z úloh. Tento článek popisuje každou společnou úlohu, data a hodnocení. metriky a uvádí výsledky obdržených podání.
[ "Antonios Anastasopoulos", "Ondřej Bojar", "Jacob Bremerman", "Roldano Cattoni", "Maha Elbayad", "Marcello Federico", "Xutai Ma", "Satoshi Nakamura", "Matteo Negri", "Jan Niehues", "Juan Pino", "Elizabeth Salesky", "Sebastian Stüker", "Katsuhito Sudoh", "Marco Turchi", "Alex Waibel", "Changhan Wang", "Matthew Wiesner" ]
https://www.semanticscholar.org/paper/0a53f5ef9769b2f08ceee72e23312974a7cdf7e7/
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
Optional valency complementations in Czech light verb constructions
Fakultativní volná doplnění v konstrukcích s kategoriálním slovesem v češtině
This paper addresses Czech light verb constructions, partly revising principles of their syntactic structure formation formulated within the Functional Generative Description. It argues that obligatoriness of valency complementations should be reflected in these principles. Namely, the role of optional valency complementations of light verbs played in this process has been analyzed. This analysis has shown that in the cases where light verbs do not provide a sufficient number of valency complementations for the surface expression of semantic participants of predicative nouns, semantic participants of nouns make use of optional verbal complementations. In such cases, semantic participants can be expressed on the surface, either as optional verbal complementation or as nominal complementation. The distribution of verbal and nominal complementations have been observed in 1,600 light verb constructions extracted from the Czech National Corpus, with the result that the surface expression of these participants through the optional verbal complementations is strongly preferred (88% of verbal complementations and 12% of nominal ones).
Příspěvek se zabývá konstrukcemi s kategoriálním slovesem v češtině. Částěčně reviduje pravidla pro formování syntaktické struktury těchto konstrukcí navržená v rámci funkčního generativního popisu, a to zejména s ohledem na roli fakultativních volných doplnění kategoriálních sloves v těchto konstrukcích. Provedená analýza ukazuje, že v případech, kdy kategoriální slovesa neposkytují dostatečný počet valenčních doplnění pro povrchové vyjádření sémantických participantů jména, využívají participanty pro své povrchové vyjádření i fakultativní volná doplnění slovesa. Distribuce slovesných a jmenných doplnění analyzovaná v 1 600 konstrukcích s kategoriálním slovesem ukázala, že povrchové vyjádření participantů jména skrze fakultativní volná doplnění kategoriálního slovesa je v češtině silně preferováno (88% fakultativních volných doplnění slovesa oproti 12% valenčních doplnění jména).
[ "Václava Kettnerová" ]
https://www.semanticscholar.org/paper/481398b63e6f53aadff1e00c03ea61388da1e266/
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
Reflexives in the VALLEX Lexicon: Syntactic Reflexivity and Reciprocity
Reflexiva ve slovníku VALLEX: Syntaktická reflexivita a reciprocita
Reflexives, encoding a variety of meanings, pose a great challenge for both theoretical and lexicographic description. As they are associated with changes in morphosyntactic properties of verbs, their description is highly relevant for verb valency. In Czech, reflexives function as the reflexive personal pronoun and as verbal affixes. In this paper, we address those language phenomena that are encoded by the reflexive personal pronoun, i.e., reflexivity and reciprocity. We introduce the lexicographic representation of these two language phenomena in the VALLEX lexicon, a valency lexicon of Czech verbs, accounting for the role of the reflexives with respect to the valency structure of verbs. This representation makes use of the division of the lexicon into a data component and a grammar component. It takes into account that reflexivity and reciprocity are conditioned by the semantic properties of verbs on the one hand and that morphosyntactic changes brought about by these phenomena are systemic on the other. About one third of the lexical units contained in the data component of the lexicon are assigned the information on reflexivity and/or reciprocity in the form of pairs of the affected valency complementations (2,039 on reflexivity and 2,744 on reciprocity). A set of rules is formulated in the grammar component (3 rules for reflexivity and 18 rules for reciprocity). These rules derive the valency frames underlying syntactically reflexive and reciprocal constructions from the valency frames describing non-reflexive and non-reciprocal constructions. Finally, the proposed representation makes it possible to determine which lexical units of verbs create ambiguous constructions that can be interpreted either as reflexive or as reciprocal.
Reflexiva představují velkou výzvu pro teoretický i lexikografický popis jazyka. Vzhledem ke změnám morfo-syntaktických vlastností sloves, která s sebou užití reflexiv nese, je jejich zachycení a popis vysoce relevantní pro slovesnou valenci. V češtině fungují reflexiva buď jako reflexivní osobní zájmeno, nebo jako morfém tvořící slovesné lemma či slovesný tvar. V tomto příspěvku se zabýváme těmi jazykovými jevy, které jsou kódovány reflexivním osobním zájmenem, tj. reflexivitou a reciprocitou. Lexikografické znázornění těchto dvou jazykových jevů zpracováváme v rámci Valenčního slovníku českých sloves VALLEX. Reprezentace ve VALLEXu využívá rozdělení lexikonu na datovou a gramatickou komponentu; bere v úvahu, že reflexivita i reciprocita jsou podmíněny sémantickými vlastnostmi sloves, ovšem morfologické změny vyvolané těmito jevy jsou systémové a lze je popsat pomocí pravidel. Zhruba třetina lexikálních jednotek obsažených v datové komponentě lexikonu má přiřazenu informaci o možném užití v reflexivních a/nebo recipročních konstrukcí, a to v podobě párů dotčených valenčních komplementů (2 039 u reflexivity a 2 744 u reciprocity). Dále je v gramatické komponentě formulován soubor pravidel pro odvození valenčních rámců pro syntakticky reflexní a reciproční konstrukce, která se aplikují na základové valenční rámce popisující nereflexní a nereferenční konstrukce (3 pravidla pro reflexivitu a 18 pravidel pro reciprocitu). V příspěvku se dále diskutují situace, kdy lexikální jednotky sloves vytvářejí nejednoznačné konstrukce, které mohou být vykládány buď jako reflexivní, nebo jako reciproční.
[ "Václava Kettnerová", "Markéta Lopatková", "Anna Vernerová" ]
https://www.semanticscholar.org/paper/f7ec8415b1e51d3dd8df89cd4b230077937a0f8e/
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
EUNIC AI Science Café Series: AI & Theatre
EUNIC AI Science Café Series: AI & divadlo
Over the past decades an increasing number of performing artists and technologists have brought robots to the center stage in their performances. AI has taken roles such as comedy improvisers, storytellers, actors, dancers, and choreographers, disrupting the traditional two-way human actor-human audience interaction. By presenting this talk we try to challenge the idea that live performance is a specifically human activity and explore different forms of human-AI interactions in the theatre. What story does AI tell? What emotions can it generate? Do you think artificial intelligence is able to create an enjoyable theatre script? Can a robot become a playwright? Does the perfect theater evening exist, produced by an autonomous machine?
V posledních desetiletích stále větší počet vystupujících umělců a technologů přivedl roboty na středové pódium při jejich vystoupeních. AI si přisvojila role komediálních improvizátorů, vypravěčů, herců, tanečníků a choreografů, čímž narušila tradiční obousměrnou lidskou herecko-lidskou diváckou interakci. Předložením této přednášky se snažíme zpochybnit myšlenku, že živé představení je specificky lidskou činností, a prozkoumat různé formy interakce mezi člověkem a umělou inteligencí v divadle. Jaký příběh Al vypráví? Jaké emoce to může vyvolat? Myslíte si, že umělá inteligence je schopna vytvořit příjemný divadelní scénář? Může se robot stát dramatikem? Existuje dokonalý divadelní večer, který vyprodukuje autonomní stroj?
[ "Tracy Harwood", "Piotr Mirowski", "Marcel Karnapke", "Rudolf Rosa" ]
https://www.semanticscholar.org/paper/586e10a5f7371f27bfd176addff014bf01401319/
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
Report on the 6th Workshop on Search-Oriented Conversational AI (SCAI 2021)
Zpráva o 6. workshopu na téma konverzační umělá inteligence zaměřená na vyhledávání
The 6th edition of the Search-Oriented Conversational AI workshop (SCAI 2021) was organised as a discussion platform on conversational AI for intelligent information access. The workshop was designed to be multidisciplinary, bringing together researchers and practitioners across the fields of natural language processing (NLP), information retrieval (IR), machine learning (ML) and human-computer interaction (HCI). The workshop included four sessions featuring invited talks, a separate poster session, and a session discussing the results of a shared task on conversational question answering (SCAI-QReCC).
Šestý ročník workshopu Search-Oriented Conversational AI (SCAI 2021) byl uspořádán jako diskusní platforma o konverzační umělé inteligenci pro inteligentní přístup k informacím. Workshop byl koncipován jako multidisciplinární a spojil výzkumné pracovníky a odborníky z praxe napříč obory zpracování přirozeného jazyka (NLP), vyhledávání informací (IR), strojového učení (ML) a interakce člověk-počítač (HCI). Workshop zahrnoval čtyři sekce, na nichž zazněly pozvané přednášky, samostatnou posterovou sekci a sekci, na níž se diskutovalo o výsledcích soutěže v konverzačním zodpovídání otázek (SCAI-QReCC).
[ "Svitlana Vakulenko", "Ondřej Dušek", "Leigh Clark" ]
https://www.semanticscholar.org/paper/6b04ae8491a3c621e1ebd9dc22ad9b4d35570b96/
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
Defining And Detecting Inconsistent System Behavior in Task-oriented Dialogues
Definice a detekce nekonzistentního chování systému v asistivních dialozích
We present experiments on automatically detecting inconsistent behavior of task-oriented dialogue systems from the context. We enrich the bAbI/DSTC2 data (Bordes et al., 2017) with automatic annotation of dialogue inconsistencies, and we demonstrate that inconsistencies correlate with failed dialogues. We hypothesize that using a limited dialogue history and predicting the next user turn can improve inconsistency classification. While both hypotheses are confirmed for a memory-networks-based dialogue model, it does not hold for a training based on the GPT-2 language model, which benefits most from using full dialogue history and achieves a 0.99 accuracy score.
Představujeme experimenty s automatickým odhalováním nekonzistentního chování na základě kontextu u dialogových systémů orientovaných na úkoly. Obohacujeme data bAbI/DSTC2 (Bordes et al., 2017) o automatickou anotaci nekonzistencí v dialogu a ukazujeme, že nekonzistence korelují s neúspěšnými dialogy. Předpokládáme, že použití omezené historie dialogů a předvídání dalšího tahu uživatele může zlepšit klasifikaci nekonzistencí. Zatímco obě hypotézy se potvrzují pro dialogový model založený na memory networks, neplatí pro trénování jazykového modelu GPT-2, který nejvíce těží z použití úplné historie dialogu a dosahuje 99% přesnosti.
[ "Léon-Paul Schaub", "Vojtěch Hudeček", "Daniel Štancl", "Ondřej Dušek", "Patrick Paroubek" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
AuGPT: Auxiliary Tasks and Data Augmentation for End-To-End Dialogue with Pre-Trained Language Models
AuGPT: Pomocné úlohy a augmentace dat pro end-to-end dialog s předtrénovanými jazykovými modely
Attention-based pre-trained language models such as GPT-2 brought considerable progress to end-to-end dialogue modelling. However, they also present considerable risks for task-oriented dialogue, such as lack of knowledge grounding or diversity. To address these issues, we introduce modified training objectives for language model finetuning, and we employ massive data augmentation via back-translation to increase the diversity of the training data. We further examine the possibilities of combining data from multiples sources to improve performance on the target dataset. We carefully evaluate our contributions with both human and automatic methods. Our model substantially outperforms the baseline on the MultiWOZ data and shows competitive performance with state of the art in both automatic and human evaluation.
Předtrénované jazykové modely založené na attention, jako je GPT-2, přinesly značný pokrok v modelování dialogů "end-to-end". Pro dialog zaměřený na úkoly však představují také značná rizika, jako je nedostatečná podloženost fakty nebo rozmanitost odpovědí. Abychom tyto problémy vyřešili, zavádíme modifikované trénovací cíle pro dotrénování jazykového modelu a využíváme masivní augmentaci dat pomocí zpětného překladu, abychom zvýšili rozmanitost trénovacích dat. Dále zkoumáme možnosti kombinace dat z více zdrojů s cílem zlepšit výkonnost na cílové datové sadě. Naše příspěvky pečlivě vyhodnocujeme pomocí lidských i automatických metod. Náš model podstatně překonává základní model na datech MultiWOZ a vykazuje výkon konkurenceschopný se současným stavem techniky při automatickém i lidském hodnocení.
[ "Jonáš Kulhánek", "Vojtěch Hudeček", "Tomáš Nekvinda", "Ondřej Dušek" ]
https://www.semanticscholar.org/paper/3ebd251e5307e91adc009c0515ea5c8e3ef44344/
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
AuGPT: Dialogue with Pre-trained Language Models and Data Augmentation
AuGPT: Dialog s předtrénovanými jazykovými modely a augmentací dat
Attention-based pre-trained language models such as GPT-2 brought considerable progress to end-to-end dialogue modelling. However, they also present considerable risks for task-oriented dialogue, such as lack of knowledge grounding or diversity. To address these issues, we introduce modified training objectives for language model finetuning, and we employ massive data augmentation via back-translation to increase the diversity of the training data. We further examine the possibilities of combining data from multiples sources to improve performance on the target dataset. We carefully evaluate our contributions with both human and automatic methods. Our model achieves state-of-the-art performance on the MultiWOZ data and shows competitive performance in human evaluation.
Předtrénované jazykové modely založené na attention, jako je GPT-2, přinesly značný pokrok v end-to-end modelování dialogů. Pro dialog zaměřený na úkoly však představují také značná rizika, jako je nedostatečná korespondence s databází nebo nedostatek rozmanitosti odpovědí. Abychom tyto problémy vyřešili, zavádíme pro doladění jazykového modelu modifikované trénovací cíle a využíváme masivní rozšíření trénovacích dat pomocí zpětného překladu, čímž zvyšujeme jejich rozmanitost. Dále zkoumáme možnosti kombinace dat z více zdrojů s cílem zlepšit výkonnost na cílové datové sadě. Naše příspěvky pečlivě vyhodnocujeme pomocí ručních i automatických metod. Náš model dosahuje nejlepších výsledků na datové sadě MultiWOZ a vykazuje konkurenceschopný výkon při lidském hodnocení.
[ "Jonáš Kulhánek", "Vojtěch Hudeček", "Tomáš Nekvinda", "Ondřej Dušek" ]
https://www.semanticscholar.org/paper/87424eafa76cbb071eca5283bd5d6e3d81dbf685/
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
The Reality of Multi-Lingual Machine Translation
Realita mnohojazyčného strojového překladu
Our book "The Reality of Multi-Lingual Machine Translation" discusses the benefits and perils of using more than two languages in machine translation systems. While focused on the particular task of sequence-to-sequence processing and multi-task learning, the book targets somewhat beyond the area of natural language processing. Machine translation is for us a prime example of deep learning applications where human skills and learning capabilities are taken as a benchmark that many try to match and surpass. We document that some of the gains observed in multi-lingual translation may result from simpler effects than the assumed cross-lingual transfer of knowledge. In the first, rather general part, the book will lead you through the motivation for multi-linguality, the versatility of deep neural networks especially in sequence-to-sequence tasks to complications of this learning. We conclude the general part with warnings against too optimistic and unjustified explanations of the gains that neural networks demonstrate. In the second part, we fully delve into multi-lingual models, with a particularly careful examination of transfer learning as one of the more straightforward approaches utilizing additional languages. The recent multi-lingual techniques, including massive models, are surveyed and practical aspects of deploying systems for many languages are discussed. The conclusion highlights the open problem of machine understanding and reminds of two ethical aspects of building large-scale models: the inclusivity of research and its ecological trace.
Naše kniha „The Reality of Multi-Lingual Machine Translation“ pojednává o výhodách a nebezpečích používání více než dvou jazyků v systémech strojového překladu. I když se kniha zaměřuje na konkrétní úkol zpracování sekvencí a víceúkolového učení (multi-task learning), cílí i poněkud mimo oblast zpracování přirozeného jazyka. Strojový překlad je pro nás ukázkovým příkladem aplikací hlubokého učení, kde jsou lidské dovednosti a schopnosti učení brány jako laťka, kterou se mnozí snaží dosáhnout a překonat. Dokumentujeme, že některé z výdobytků pozorovaných v mnohojazyčném překladu mohou vyplývat z jednodušších důvodů, než je předpokládaný přenos znalostí napříč jazyky. V první, spíše obecné části vás kniha provede motivací pro mnohojazyčnost, univerzálností hlubokých neuronových sítí zejména v úlohách typu sequence-to-sequence až ke komplikacím tohoto učení. Obecnou část uzavíráme varováním před příliš optimistickým a neopodstatněným vysvětlením zlepšení, které neuronové sítě demonstrují. Ve druhé části se plně ponoříme do mnohojazyčných modelů, se zvlášť pečlivým zkoumáním učení přenosem (transfer learning) jako jednoho z přímočařejších přístupů využívajících další jazyky. Zkoumány jsou současné vícejazyčné techniky, včetně masivních modelů, a diskutovány jsou praktické aspekty nasazení mnohojazyčných systémů. Závěr knihy zdůrazňuje otevřený problém strojového porozumění a připomíná dva etické aspekty budování rozsáhlých modelů: inkluzivnost výzkumu a jeho ekologický dopad.
[ "Tom Kocmi", "Dominik Macháček", "Ondřej Bojar" ]
https://www.semanticscholar.org/paper/be9cb71d965ca144ae8329d0ed8bd8b27dcb740b/
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
Surprise Language Challenge: Developing a Neural Machine Translation System between Pashto and English in Two Months
Jazyková výzva s překvapením: Vývoj systému neurálního strojového překladu mezi paštštinou a angličtinou za dva měsíce
In the media industry, the focus of global reporting can shift overnight. There is a compelling need to be able to develop new machine translation systems in a short period of time, in order to more efficiently cover quickly developing stories. As part of the low-resource machine translation project GoURMET, we selected a surprise language for which a system had to be built and evaluated in two months (February and March 2021). The language selected was Pashto, an Indo-Iranian language spoken in Afghanistan, Pakistan and India. In this period we completed the full pipeline of development of a neural machine translation system: data crawling, cleaning, aligning, creating test sets, developing and testing models, and delivering them to the user partners. In this paper we describe the rapid data creation process, and experiments with transfer learning and pretraining for Pashto-English. We find that starting from an existing large model pre-trained on 50 languages leads to far better BLEU scores than pretraining on one high-resource language pair with a smaller model. We also present human evaluation of our systems, which indicates that the resulting systems perform better than a freely available commercial system when translating from English into Pashto direction, and similarly when translating from Pashto into English.
V mediálním průmyslu se zaměření globálního zpravodajství může přes noc změnit. Existuje přesvědčivá potřeba být schopni vyvinout nové systémy strojového překladu v krátkém časovém období, aby bylo možné efektivněji pokrýt rychle se vyvíjející příběhy. Jako součást stroje s nízkými zdroji překladatelského projektu GOURMET jsme náhodně vybrali jazyk, pro který musel být systém postaveno a vyhodnoceno za dva měsíce (únor a březen 2021). Vybraný jazyk byl Paštština, indoíránský jazyk používaný v Afghánistánu, Pákistánu a Indii. V tomto období jsme dokončili celý proces vývoje systému neuronového strojového překladu: procházení dat, čištění, zarovnání, vytváření testovacích sad, vývoj a testování modelů a jejich poskytování uživatelským partnerům. V tomto článku popisujeme rychlý proces vytváření dat a experimenty s transferovým učením a přípravou na paštskou angličtinu. Zjišťujeme, že začínáme od existujícího velký model předem proškolený na 50 jazycích vede k mnohem lepším výsledkům BLEU než předtrénování na jeden vysoce zdrojový jazykový pár s menším modelem. Uvádíme také lidské hodnocení naše systémy, což naznačuje, že výsledné systémy fungují lépe než volně dostupné komerční systém při překladu z angličtiny do paštštiny a podobně při překlad z paštštiny do angličtiny.
[ "Alexandra Birch", "Barry Haddow", "Antonio Valerio Miceli Barone", "Jindřich Helcl", "Jonas Waldendorf", "Felipe Sánchez-Martínez", "Mikel L. Forcada", "Víctor M. Sánchez-Cartagena", "Juan Antonio Pérez-Ortiz", "Miquel Esplà-Gomis", "Wilker Aziz", "Lina Murady", "Sevi Sariisik", "Peggy van der Kreeft", "Kay MacQuarrie" ]
https://www.semanticscholar.org/paper/ca2cbf18b53516093fb3f0705121b9d63c5c34ed/
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
A Corpus with Wavesurfer and TEI: Speech and Video in TEITOK
Korpus s Wavesurferem a TEI: Řeč a video v TEITOK
In this paper, we demonstrate how TEITOK provides a full online interface for speech and even video corpora, that are fully searchable using the CQL query language, can contain all speech-related annotation such as repetitions, gaps, and mispronunciations, and provides a full interface for time-aligned annotations scrolling below the waveform and showing the video if there is any. Corpora are stored in the TEI/XML standard, with import and output functions for other established standards like ELAN, Praat, or Transcriber. It is even possible to directly annotate corpora in TEITOK.
V tomto článku demonstrujeme, jak TEITOK poskytuje úplné online rozhraní pro řeč a dokonce i video korpusy, které jsou plně prohledávatelné pomocí dotazovacího jazyka CQL, mohou obsahovat všechny anotace související s řečí, jako jsou opakování, mezery a nesprávné výslovnosti, a poskytuje úplnou rozhraní pro časově zarovnané anotace rolující pod křivkou a zobrazení videa, pokud nějaké existuje. Korpusy jsou uloženy ve standardu TEI/XML s funkcemi importu a výstupu pro další zavedené standardy, jako je ELAN, Praat nebo Transcriber. V TEITOK je dokonce možné korpusy přímo anotovat.
[ "Maarten Janssen" ]
https://www.semanticscholar.org/paper/b1018746b1251a4b80679393e7d151c6214bebd8/
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
Integrating TEITOK and KonText/PMLTQ at LINDAT
Integrace TEITOK a KonText/PMLTQ na LINDAT
In this paper we describe how the TEITOK corpus platform was integrated with the KonText and PML-TQ corpus platforms at LINDAT to provide document visualization for both existing and future resources at LINDAT. TEITOK is an online platform for searching, viewing, and editing corpora, where corpus files are stored as annotated TEI/XML files. The TEITOK integration also means LINDAT resources will become available in TEI/XML format, and searchable in CWB on top of existing tools at the institute. Although the integration described in this paper is specific for LINDAT, the method should be applicable to the integration of TEITOK or similar tools into an existing corpus architecture.
V tomto příspěvku popisujeme, jak byla korpusová platforma TEITOK integrována s korpusovými platformami KonText a PML-TQ v LINDAT, aby poskytla vizualizaci dokumentů pro stávající i budoucí zdroje v LINDAT. TEITOK je online platforma pro vyhledávání, prohlížení a úpravy korpusů, kde jsou soubory korpusu ukládány jako anotované soubory TEI / XML. Integrace TEITOK také znamená, že zdroje LINDAT budou k dispozici ve formátu TEI / XML a lze je vyhledávat v CWB nad rámec stávajících nástrojů ústavu. Ačkoli integrace popsaná v tomto článku je specifická pro LINDAT, tato metoda by měla být použitelná pro integraci TEITOK nebo podobných nástrojů do existující korpusové architektury.
[ "Maarten Janssen" ]
https://www.semanticscholar.org/paper/605a4ffa131cad04dce63bfe3c6dedcaefc68a94/
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
UDWiki: guided creation and exploitation of UD treebanks
UDWiki: řízená tvorba a využívání UD stromových bank
UDWiki is an online environment designed to make creating new UD treebanks easier. It helps in setting up all the necessary data needed for a new treebank up in a GUI, where the interface takes care of guiding you through all the descriptive files needed, adding new texts to your corpus, and helping in annotating the texts. The system is built on top of the TEITOK corpus environment, using an XML based version of UD annotation, where dependencies can be combined with various other types of annotations. UDWiki can run all the necessary or helpful scripts (taggers, parsers, validators) via the interface. It also makes treebanks under development directly searchable, and can be used to maintain or search existing UD treebanks.
UDWiki je online prostředí navržené tak, aby usnadnilo vytváření nových stromových bank UD. Pomáhá při nastavení všech nezbytných dat potřebných pro novou stromovou banku v GUI, kde se rozhraní postará o to, aby vás provedlo všemi potřebnými popisnými soubory, přidalo nové texty do vašeho korpusu a pomohlo s anotací textů. Systém je postaven na korpusovém prostředí TEITOK a využívá verzi UD anotací na bázi XML, kde lze závislosti kombinovat s různými jinými typy anotací. UDWiki může spouštět všechny potřebné nebo užitečné skripty (tagery, parsery, validátory) přes rozhraní. Umožňuje také přímé vyhledávání ve vyvíjených stromových bankách a lze je použít k údržbě nebo prohledávání stávajících stromových bank UD.
[ "Maarten Janssen" ]
https://www.semanticscholar.org/paper/dd62869425aedfec25a75f084eea76556db499da/
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
Deep Universal Dependencies 2.7
Deep Universal Dependencies 2.7
Deep Universal Dependencies is a collection of treebanks derived semi-automatically from Universal Dependencies. It contains additional deep-syntactic and semantic annotations. Version of Deep UD corresponds to the version of UD it is based on. Note however that some UD treebanks have been omitted from Deep UD.
Deep Universal Dependencies (hluboké univerzální závislosti) je sbírka treebanků poloautomaticky odvozených z Universal Dependencies. Obsahuje přídavné hloubkově-syntaktické a sémantické anotace. Verze Deep UD odpovídá verzi UD, na které je založena. Mějte však na paměti, že některé treebanky UD nebyly do Deep UD zahrnuty.
[ "Kira Droganova", "Daniel Zeman" ]
https://www.semanticscholar.org/paper/a2d25cf45f034108628a044b5470ed358164fc02/
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
The GEM Benchmark: Natural Language Generation, its Evaluation and Metrics
Benchmark GEM: Generování přirozeného jazyka, jeho evaluace a metriky
We introduce GEM, a living benchmark for natural language Generation (NLG), its Evaluation, and Metrics. Measuring progress in NLG relies on a constantly evolving ecosystem of automated metrics, datasets, and human evaluation standards. Due to this moving target, new models often still evaluate on divergent anglo-centric corpora with well-established, but flawed, metrics. This disconnect makes it challenging to identify the limitations of current models and opportunities for progress. Addressing this limitation, GEM provides an environment in which models can easily be applied to a wide set of tasks and in which evaluation strategies can be tested. Regular updates to the benchmark will help NLG research become more multilingual and evolve the challenge alongside models. This paper serves as the description of the data for the 2021 shared task at the associated GEM Workshop.
Představujeme GEM, živý benchmark pro generování přirozeného jazyka (NLG), jeho evaluaci a metriky. Měření pokroku v oblasti NLG se opírá o neustále se vyvíjející ekosystém automatizovaných metrik, datových sad a standardů lidské evaluace. Vzhledem k tomuto pohyblivému cíli se nové modely často stále vyhodnocují na odlišných anglocentrických korpusech s dobře zavedenými, ale chybnými metrikami. Tato nesouvislost ztěžuje identifikaci omezení současných modelů a příležitostí k pokroku. GEM toto omezení řeší a poskytuje prostředí, v němž lze modely snadno aplikovat na široký soubor úloh a v němž lze testovat evaluační strategie. Pravidelné aktualizace benchmarku pomohou výzkumu NLG stát se více mnohojazyčným a rozvíjet úlohu spolu s modely. Tento článek slouží jako popis dat pro sdílenou úlohu 2021 na souvisejícím workshopu GEM.
[ "Sebastian Gehrmann", "Tosin Adewumi", "Karmanya Aggarwal", "Pawan Sasanka Ammanamanchi", "Aremu Anuoluwapo", "Antoine Bosselut", "Kyathi Raghavi Chandu", "Miruna Clinciu", "Dipanjan Das", "Kaustubh D. Dhole", "Wanyu Du", "Esin Durmus", "Ondřej Dušek", "Chris Emezue", "Varun Gangal", "Cristina Garbacea", "Tatsunori Hashimoto", "Yufang Hou", "Yacine Yernite", "Harsh Jhamtani", "Yangfeng Ji", "Shailza Jolly", "Mihir Kale", "Dhruv Kumar", "Faisal Ladhak", "Aman Madaan", "Mounica Maddela", "Khyati Mahajan", "Saad Mahamood", "Bodhisattwa Prasad Majumder", "Pedro Henrique Martins", "Angelina McMillan-Major", "Simon Mille", "Emiel van Miltenburg", "Moin Nadeem", "Shashi Narayan", "Vitaly Nikolaev", "Rubungo Andre Niyongabo", "Salomey Osei", "Ankur Parikh", "Laura Perez-Beltrachini", "Niranjan Ramesh Rao", "Vikas Raunak", "Juan Diego Rodriguez", "Sashank Santhanam", "Joao Sedoc", "Thibault Sellam", "Samira Shaikh", "Anastasia Shimorina", "Marco Antonio Sobrevilla Cabezudo", "Hendrik Strobelt", "Nishant Subramani", "Wei Xu", "Diyi Yang", "Akhila Yerukola", "Jiawei Zhou" ]
https://www.semanticscholar.org/paper/824cd8db8a68732db04f4d8b7139eb4475e59ff2/
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
"Getting on the page": The practical accord of material resources in educational interaction
Praktický soulad materiálních prostředků ve vzdělávací interakci
This article explicates the notion of ‘practical accord’, encompassing the routine nonproblematic relationship of material artifacts and textual objects that provides the necessary grounds for further classroom work. Practical accord consists of courses of action that are temporally aligned with structured objects such as series of pages, slides or questions, ordered as a sequence of steps. Grounded in the video-based analysis of a single case, the article argues that such practical accord can be a necessary requirement for an educative activity to take place in an orderly way in the classroom. The analysis focuses on a particular instance of three students working with a shared laptop and a paper worksheet, and losing their grasp of the relationship between the screen and the sheet. The identified practices used to get back ‘on the page’ include the verbal and gestural constitution of the screen and the sheet as two separate objects that are related through instructions provided on the screen, which serves as a link between two independent but interrelated numbering systems used to organise the on-screen material and the questions on the worksheet. The article concludes with a discussion of the notion of practical accord with regard to instructed action and gestalt contextures.
Tento článek vysvětluje pojem „praktický soulad“, zahrnující rutinní neproblematický vztah hmotných artefaktů a textových předmětů, který poskytuje nezbytný základ pro práci ve třídě. Analýza videonahrávky interakce se zaměřuje na konkrétní případ tří studentů, kteří společně pracují se sdíleným notebookem a papírovým pracovním listem.
[ "Jakub Mlynář" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
Rewatching a video clip in classroom work with digital oral history
Opakované sledování videoklipu ve třídě během práce s digitální orální historií
This paper focuses on rewatching sequences in classroom work with digital oral history, i.e., participants conducting a second watching of a video clip that they have already seen. I provide a detailed examination of how the participants initiate rewatching using speech and embodied conduct. They treat the video clip as a structured object containing an already familiar succession of utterances by the narrator. The rewatching sequences are closed when the participants have identified a particular utterance within the video that they can treat as a solution to a previously encountered ambiguity. To do this, they search for the relevant utterances by manipulating the time marker on the screen and stabilize them by pausing the clip. The participants thus exhibit a local temporary competence in rewatching and manipulating video clips as topics and resources in classroom tasks.
Článek se zaměřuje na sekvence opětovného sledování ve školní práci s digitální orální historií, tj. na účastníky provádějící druhé sledování videoklipu, který již jednou viděli. Podrobně zkoumám, jak účastníci zahajují přehrávání pomocí řeči a ztělesněného chování. K videoklipu přistupují jako ke strukturovanému objektu obsahujícímu již známý sled promluv vypravěče. Sekvence opakovaného sledování se uzavřou, když účastníci v rámci videa naleznou konkrétní promluvu, který mohou považovat za řešení dříve zjištěné nejednoznačnosti. K tomu vyhledávají relevantní promluvy manipulací s časovou značkou na obrazovce a stabilizují je zastavením klipu. Účastníci tak vykazují místní dočasnou kompetenci opětovně sledovat videoklipy a manipulovat s nimi jako s relevantními tématy a zdroji při řešení úkolů ve školní třídě.
[ "Jakub Mlynář" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
cs
2021
Discussion after the performance AI: When a Robot Writes a Play
Diskuze po představení AI: Když robot píše hru
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.
[ "Martina Kinská", "Rudolf Rosa" ]
https://www.semanticscholar.org/paper/9e3e68a22a61903bee336d88b3d986e789ca17d2/
null
null
null
null
null
null
null
null
null
null
null
null
cs
2021
AI: When a Robot Writes a Play (theatre programme)
AI: Když robot píše hru (divadelní program)
On the occasion of the centenary of Karel Čapek's play R.U.R, the Švanda Theatre prepared a presentation of a unique project THEAITRE that examines whether artificial intelligence can write a play. Within a few months, the computer generated images from the life of a robot that has to face the joys and sorrows of everyday life. And it revealed to us how it perceives basic human issues such as birth, dying, the desire for love, the search for jobs, or aging.
Švandovo divadlo při příležitosti stoletého výročí premiéry hry R.U.R Karla Čapka uvádí prezentaci unikátního projektu THEAITRE, který zkoumá, zda umělá inteligence dokáže napsat divadelní hru. Během několika měsíců počítač generoval obrazy ze života robota, který musí čelit radostem a strastem každodenního života. A odkryl nám, jak vnímá základní lidské otázky jako je narození, umírání, touha po lásce, hledání pracovních příležitostí, či stárnutí.
[ "Martina Kinská", "David Košťák", "Ondřej Dušek", "Rudolf Rosa" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
Transferring Word-Formation Networks Between Languages
Přenos slovotvorných sítí mezi jazyky
In this article, we present a proof-of-concept method for creating word-formation networks by transferring information from another language. The proposed algorithm utilizes an existing word-formation network and parallel texts and creates a low-precision and moderate-recall network in a language, for which no manual annotations need to be available. We then extend the coverage of the resulting network by using it to train a machine-learning method and applying the resulting model to a larger lexicon, obtaining a moderate-precision and high-recall result. The approach is evaluated on French, German and Czech against existing word-formation networks in those languages.
V tomto článku představujeme metodu pro vytváření slovotvorných sítí pomocí přenosu informací z jiného jazyka. Navrhovaný algoritmus využívá existující slovotvornou síť a paralelní texty a vytváří síť s nízkou precision a středním recallem v jazyce, pro který nemusí být k dispozici manuální anotace. Recall výsledné sítě pak rozšířujeme tím, že ji využijeme k natrénování metody strojového učení a výsledný model aplikujeme na větší slovník, čímž získáme výsledek s obdobnou precision, ale vyšším recallem. Přístup je vyhodnocován proti existujícím slovotvorným sítím ve francouzštině, němčině a češtině.
[ "Jonáš Vidra", "Zdeněk Žabokrtský" ]
https://www.semanticscholar.org/paper/aefd413693b2b19cb7234ddd69a2d49826a06ede/
null
null
null
null
null
null
null
null
null
null
null
null
cs
2021
DeriNet v2.1
DeriNet v2.1
DeriNet is a lexical network which models derivational relations in the lexicon of Czech. Nodes of the network correspond to Czech lexemes, while edges represent word-formational relations between a derived word and its base word / words. The present version, DeriNet 2.1, contains 1,039,012 lexemes (sampled from the MorfFlex CZ 2.0 ​dictionary) connected by 782,814 derivational, 50,533 orthographic variant, 1,952 compounding, 295 univerbation and 144 conversion relations. Compared to the previous version, version 2.1 contains annotations of orthographic variants, full automatically generated annotation of affix morpheme boundaries (in addition to the roots annotated in 2.0), 202 affixoid lexemes serving as bases for compounding, annotation of corpus frequency of lexemes, annotation of verbal conjugation classes and a pilot annotation of univerbation. The set of part-of-speech tags was converted to Universal POS from the Universal Dependencies project.
DeriNet je lexikální síť, která modeluje derivační vztahy ve slovníku češtiny. Uzly sítě odpovídají českým lexémům, zatímco hrany představují slovotvorné vztahy mezi odvozeným slovem a jeho základním slovem / slovy. Současná verze, DeriNet 2.1, obsahuje 1 039 012 lexemes (extrahovány ze slovníku MorfFlex CZ 2.0) spojených 782 814 derivačními relacemi, 50 533 ortografickými variantami, 1 952 vztahů skládání, 295 univerbizačními vztahy a 144 konverzními vztahy. Ve srovnání s předchozí verzí obsahuje verze 2.1 anotace ortografických variant, plně automaticky generovanou anotaci hranic afixů (kromě kořenů anotovaných v 2.0), 202 affixoidů sloužících jako základ pro skládání, anotaci četností lexémů z korpusů, anotaci slovesných tříd a pilotní anotaci univerbace. Sada tagů pro slovní druhy byla převedena na Universal POS z projektu Universal Dependencies.
[ "Jonáš Vidra", "Zdeněk Žabokrtský", "Lukáš Kyjánek", "Magda Ševčíková", "Šárka Dohnalová", "Emil Svoboda", "Jan Bodnár" ]
https://www.semanticscholar.org/paper/e24c79638d8da46532aa6feadb705f4f03a142fa/
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
Splitting and Identifying Czech Compounds: A Pilot Study
Splitting a identifikace českých kompozit: pilotní studie
We present pilot experiments on splitting and identifying Czech compound words. We created an algorithm measuring the linguistic similarity of two words based on finding the shortest path through a matrix of mutual estimated correspondences between two phonemic strings. Additionally, a neural compound-splitting tool (Czech Compound Splitter) was implemented by using the Marian Neural Machine Translator framework, which was trained on a data set containing 1,164 hand-annotated compounds and about 280,000 synthetically created compounds. In compound splitting, the first solution achieved an accuracy of 28% and the second solution achieved 54% on a separate validation data set. In compound identification, the Czech Compound Splitter achieved an accuracy of 91%.
Představujeme pilotní experimenty na dělení a identifikaci českých složených slov. Vytvořili jsme algoritmus měřící jazykovou podobnost dvou slov založený na nalezení nejkratšího cesta skrze matici vzájemných odhadovaných korespondencí mezi dvěma fonologicky přepsanými řetězci. Dále jsme vytvořili nástroj pro splitting neboli dělení složených slov (Czech Compound Splitter) pomocí frameworku Marian Neural Machine Translator, který byl vytrénován na datové sadě obsahující 1 164 ručně anotovaných sloučenin a zhruba 280 000 synteticky vytvořených kompozit. Ve splittingu kompozit dosáhlo první řešení přesnosti 28 % a druhé řešení 54 % na validačním datové sadě. V úloze identifikace kompozit dosáhl Czech Compound Splitter přesnosti 91%.
[ "Emil Svoboda", "Magda Ševčíková" ]
https://www.semanticscholar.org/paper/7f8674949bbaa7f055b410ea29655ad74f3db6d4/
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
Findings of the WMT Shared Task on Machine Translation Using Terminologies.
Závěry úlohy strojového překladu pomocí terminologií na WMT
Language domains that require very carefuluse of terminology are abundant and reflect asignificant part of the translation industry. Inthis work we introduce a benchmark for eval-uating the quality and consistency of terminol-ogy translation, focusing on the medical (andCOVID-19 specifically) domain for five lan-guage pairs: English to French, Chinese, Rus-sian, and Korean, as well as Czech to Ger-man. We report the descriptions and resultsof the participating systems, commenting onthe need for further research efforts towardsboth more adequate handling of terminologiesas well as towards a proper formulation andevaluation of the task.
Jazykové domény vyžadující velmi pečlivé používání terminologie jsou hojné a odrážejí významnou část překladatelského průmyslu. V této práci představujeme metriku pro hodnocení kvality a konzistentnosti překladu terminologie se zaměřením na lékařskou (a konkrétně COVID-19) doménu pro pět jazykových párů: angličtinu do francouzštiny, čínštiny, ruštiny a korejštiny, a dále češtinu do němčiny. Uvádíme popisy a výsledky zúčastněných systémů, vyjadřujeme nutnost dalšího výzkumu jak pro adekvátnější zacházení s terminologií, tak pro správnou formulaci a vyhodnocení úlohy.
[ "Md Mahfuz ibn Alam", "Ivana Kvapilíková", "Antonios Anastasopoulos", "Laurent Besacier", "Georgiana Dinu", "Marcello Federico", "Matthias Gallé", "Philipp Koehn", "Vassilina Nikoulina", "Kweon Woo Jung" ]
https://www.semanticscholar.org/paper/5ccde19b5d0562eac0cb67426cba9e5312badcd0/
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
Coreference meets Universal Dependencies – a pilot experiment on harmonizing coreference datasets for 11 languages
Setkání koreference a Universal Dependencies – pilotní experiment s harmonizací koreferenčních korpusů pro 11 jazyků
We describe a pilot experiment aimed at harmonizing diverse data resources that contain coreference-related annotations. We converted 17 existing datasets for 11 languages into a common annotation scheme based on Universal Dependencies, and released a subset of the resulting collection publicly under the name CorefUD 0.1 via the LINDAT-CLARIAH-CZ repository (http://hdl.handle.net/11234/1-3510).
Popisujeme pilotní experiment zaměřený na harmonizaci různorodých datových zdrojů, které obsahují anotace související s koreferencí. Převedli jsme 17 existujících korpusů 11 jazyků do společného anotačního schématu, založeného na Universal Dependencies, a zveřejnili jsme podmnožinu této kolekce pod názvem CorefUD 0.1 v repozitáři LINDAT-CLARIAH-CZ (http://hdl.handle.net/11234/1-3510).
[ "Anna Nedoluzhko", "Michal Novák", "Martin Popel", "Zdeněk Žabokrtský", "Daniel Zeman" ]
https://www.semanticscholar.org/paper/b95995b020d9ac4ab498ec9edc4969cdaf7f5837/
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
CorefUD 0.1
CorefUD 0.1
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 0.1 consists of 17 datasets for 11 languages.
CorefUD je kolekce existujících korpusů s anotací koreference, které jsme převedli na jednotné anotační schéma. Ve své současné verzi 0.1 CorefUD celkem obsahuje 17 korpusů pro 11 jazyků.
[ "Anna Nedoluzhko", "Michal Novák", "Martin Popel", "Zdeněk Žabokrtský", "Daniel Zeman" ]
https://www.semanticscholar.org/paper/33336cdc37455107ca34636d844ab352e410eb1a/
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
CorefUD 0.2
CorefUD 0.2
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 0.1, the automatic morpho-syntactic annotation has improved.
CorefUD je kolekce existujících korpusů s anotací koreference, které jsme převedli na jednotné anotační schéma. Ve své současné verzi 0.2 CorefUD celkem obsahuje 17 korpusů pro 11 jazyků a v porovnání s verzí 0.1 obsahuje kvalitnejší automatickou morfo-syntaktickou anotaci.
[ "Anna Nedoluzhko", "Michal Novák", "Martin Popel", "Zdeněk Žabokrtský", "Daniel Zeman" ]
https://www.semanticscholar.org/paper/33336cdc37455107ca34636d844ab352e410eb1a/
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
Is one head enough? Mention heads in coreference annotations compared with UD-style heads
Stačí jedna hlava? Hlavy zmínek v anotacích koreference ve srovnání s hlavami podle UD
We present an empirical study that compares mention heads as annotated manually in four coreference datasets (for Dutch, English, Polish, and Russian) on one hand, with heads induced from dependency trees parsed automatically, on the other hand.
Předkládáme empirickou studii, která srovnává hlavy zmínek anotovaných ručně ve čtyřech koreferenčních datových sadách (pro nizozemštinu, angličtinu, polštinu a ruštinu) na jedné straně a hlavy vyplývající z automaticky predikovaných závislostních syntaktických stromů na straně druhé.
[ "Anna Nedoluzhko", "Michal Novák", "Martin Popel", "Zdeněk Žabokrtský", "Daniel Zeman" ]
https://www.semanticscholar.org/paper/4229c9efe16794893d92105c77610fc2d3bb7a23/
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
ARGUABLY @ AI Debater-NLPCC 2021 Task 3: Argument Pair Extraction from Peer Review and Rebuttals
ARGUABLY @ AI Debater-NLPCC 2021 Úkol 3: Argument Párová extrakce z recenze a rozborů
This paper describes our participating system run to the argumentative text understanding shared task for AI Debater at NLPCC 2021 (http://www.fudan-disc.com/sharedtask/AIDebater21/tracks.html). The tasks are motivated towards developing an autonomous debating system. We make an initial attempt with Track-3, namely, argument pair extraction from peer review and rebuttal where we extract arguments from peer reviews and their corresponding rebuttals from author responses. Compared to the multi-task baseline by the organizers, we introduce two significant changes: (i) we use ERNIE 2.0 token embedding, which can better capture lexical, syntactic, and semantic aspects of information in the training data, (ii) we perform double attention learning to capture long-term dependencies. Our proposed model achieves the state-of-the-art results with a relative improvement of 8.81% in terms of F1 score over the baseline model. We make our code available publicly at https://github.com/guneetsk99/ArgumentMining_SharedTask. Our team ARGUABLY is one of the third prize-winning teams in Track 3 of the shared task.
Tento dokument popisuje náš systém účasti na argumentačním textu chápajícím sdílený úkol pro AI Debater na NLPCC 2021 (http://www.fudan-disc.com/sharedtask/AIDebater21/tracks.html). Úkoly jsou motivovány k rozvoji autonomního diskusního systému. Počáteční pokus provedeme s trackem-3, konkrétně extrakcí argumentačních dvojic z recenzního posudku a vyvrácením, kdy extrahujeme argumenty z recenzního posudku a jejich odpovídající vyvrácení z autorových odpovědí. Ve srovnání s víceúkolovou základnou organizátorů zavádíme dvě významné změny: i) používáme vkládání tokenu ERNIE 2.0, které dokáže lépe zachytit lexikální, syntaktické a sémantické aspekty informací v tréninkových datech, ii) provádíme dvojí učení pozornosti, abychom zachytili dlouhodobé závislosti. Náš navrhovaný model dosahuje nejmodernějších výsledků s relativním zlepšením skóre F1 o 8,81% oproti základnímu modelu. Náš kód zveřejníme na adrese https://github.com/guneetsk99/ArgumentMining_SharedTask. Náš tým ARGUABLY je jedním z třetích oceněných týmů v Tracku 3 společného úkolu.
[ "Guneet Singh Kohli", "Prabsimran Kaur", "Muskaan Singh", "Tirthankar Ghosal", "Prashant Singh Rana" ]
https://www.semanticscholar.org/paper/f46be3362c07532be1fb1e1944cbfdb88e395ad5/
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
WMT21 Marian translation model (ca-oc)
WMT21 překladový model pro systém Marian (ca-oc)
Marian NMT model for Catalan to Occitan translation. The model was submitted to WMT'21 Shared Task on Multilingual Low-Resource Translation for Indo-European Languages as a CUNI-Primary system for Catalan to Occitan.
Překladový model z katalánštiny do okcitánštiny pro systém Marian.
[ "Josef Jon" ]
https://www.semanticscholar.org/paper/2cd7bd19c9c2868a88e4c80c0c4f3cf0dd63900a/
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
WMT21 Marian translation models (ca-ro,it,oc)
WMT21 překladový model pro systém Marian (ca-ro,it,oc)
Marian multilingual translation model from Catalan into Romanian, Italian and Occitan. This model is an updated version (trained for 2.1M updates) of the model submitted to WMT'21 Shared Task on Multilingual Low-Resource Translation for Indo-European Languages as a CUNI-Primary system for Catalan to Romanian and Italian (trained for 430k updates).
Vícejazyčný překladový model z katalánštiny do rumunštiny, italštiny a okcitánštiny pro systém Marian.
[ "Josef Jon" ]
https://www.semanticscholar.org/paper/2cd7bd19c9c2868a88e4c80c0c4f3cf0dd63900a/
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
End-to-End Lexically Constrained Machine Translation for Morphologically Rich Languages
Ucelený strojový překlad s lexikálními omezeními pro mofrologicky bohaté jazyky
Lexically constrained machine translation allows the user to manipulate the output sentence by enforcing the presence or absence of certain words and phrases. Although current approaches can enforce terms to appear in the translation, they often struggle to make the constraint word form agree with the rest of the generated output. Our manual analysis shows that 46% of the errors in the output of a baseline constrained model for English to Czech translation are related to agreement. We investigate mechanisms to allow neural machine translation to infer the correct word inflection given lemmatized constraints. In particular, we focus on methods based on training the model with constraints provided as part of the input sequence. Our experiments on the English-Czech language pair show that this approach improves the translation of constrained terms in both automatic and manual evaluation by reducing errors in agreement. Our approach thus eliminates inflection errors, without introducing new errors or decreasing the overall quality of the translation.
Lexikálně omezený strojový překlad umožňuje uživateli manipulovat s výstupní větou vynucením přítomnosti nebo nepřítomnosti určitých slov a frází. Přestože současné přístupy dokáží vynutit, aby se v překladu objevily specifikované termíny, často se snaží, aby povrchová forma omezeného slova souhlasila se zbytkem vygenerovaného výstupu. Ruční analýza ukazuje, že 46% chyb ve výstupu základního omezeného modelu překladu z angličtiny do češtiny souvisí s gramatickou shodou. Zkoumáme mechanismy, které umožňují neuronových strojový překlad k určení správné inflexe omezujících slov specifikovaných pomocí lemmat. Zaměřujeme se zejména na metody založené na tréninku modelu s omezeními, které jsou součástí vstupu. Naše experimenty na anglicko-českém jazykovém páru ukazují, že tento přístup zlepšuje překlad s omezením pomocí termínů a to jak v automatickém i ručním hodnocení, snížením počtu chyb v gramatické shodě. Náš přístup tak odstraňuje inflexní chyby, aniž by zaváděl nové chyby nebo snižoval celkovou kvalitu překladu.
[ "Josef Jon", "João Paulo de Souza Aires", "Dušan Variš", "Ondřej Bojar" ]
https://www.semanticscholar.org/paper/9d030241eb35c38bacf689df6e0dfa9dadc789da/
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
CUNI systems for WMT21: Multilingual Low-Resource Translation for Indo-European Languages Shared Task
Systémy UK pro WMT21: Vícejazyčný překlad indoeurópskych jazyků s nedostatečnými zdroji
This paper describes Charles University submission for Multilingual Low-Resource Translation for Indo-European Languages shared task at WMT21. We competed in translation from Catalan into Romanian, Italian and Occitan. Our systems are based on shared multilingual model. We show that using joint model for multiple similar language pairs improves upon translation quality in each pair. We also demonstrate that chararacter-level bilingual models are competitive for very similar language pairs (Catalan-Occitan) but less so for more distant pairs. We also describe our experiments with multi-task learning, where aside from a textual translation, the models are also trained to perform grapheme-to-phoneme conversion.
Článek popisuje příspěvek Univerzity Karlovy do soutěže ve vícejazyčném překladu indoeurópskych jazyků s nedostatečnými zdroji na konferenci WMT21.
[ "Josef Jon", "Michal Novák", "João Paulo de Souza Aires", "Dušan Variš", "Ondřej Bojar" ]
https://www.semanticscholar.org/paper/41f9154e3027d1056381f994b04975960aa14b68/
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
CUNI systems for WMT21: Terminology translation Shared Task
Systémy UK pro WMT21: Překlad terminologie
This paper describes Charles University submission for Terminology translation Shared Task at WMT21. The objective of this task is to design a system which translates certain terms based on a provided terminology database, while preserving high overall translation quality. We competed in English-French language pair. Our approach is based on providing the desired translations alongside the input sentence and training the model to use these provided terms. We lemmatize the terms both during the training and inference, to allow the model to learn how to produce correct surface forms of the words, when they differ from the forms provided in the terminology database. Our submission ranked second in Exact Match metric which evaluates the ability of the model to produce desired terms in the translation.
Článek popisuje příspěvek Univerzity Karlovy do soutěže v překladu terminologie na konferenci WMT21.
[ "Josef Jon", "Michal Novák", "João Paulo de Souza Aires", "Dušan Variš", "Ondřej Bojar" ]
https://www.semanticscholar.org/paper/3d7da3c4e7a8d76b00cfb365f9a099f88740204e/
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
CUNI-MTIR at COVID-19 MLIA @ Eval Task 3
Tým CUNI-MTIR v soutěži COVID-19 MLIA @ Eval Task 3
This paper describes the participation of our team (CUNIMTIR) in the COVID-19 MILA Machine Translation (MT) task. We present our implementation of four systems (English into French, German, Swedish and Spanish) in both constrained and unconstrained settings. We employ the Marian implementation of the Transformer model to train the constrained systems in the given training data (MLIA MT parallel data), while in the unconstrained systems, we use external medicaldomain training data for the base models and then fine-tune those models using MLIA MT data
Tento článek popisuje účast týmu CUNI-MTIR v soutěži COVID-19 MLIA @ Eval Task 3.
[ "Shadi Saleh", "Hashem Sellat", "Hadi Abdi Khojasteh", "Pavel Pecina" ]
https://www.semanticscholar.org/paper/80a0c9a8ec5b23bc0820533809ff460ef9361fe5/
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
CUNI-MTIR at COVID-19 MLIA @ Eval Task 2
Tým CUNI-MTIR v soutěži COVID-19 MLIA @ Eval Task 2
The multi-lingual search task in MLIA Community Effort aims at improving COVID-19 related information access for searchers in multi-lingual settings [3]. We choose in our participation to build a monolingual system where we index the provided English documents and use the English queries for retrieval (monolingual system) then we design five runs in the monolingual settings. As for the bilingual task, we design five runs where the documents are in English, and the queries are translated into English following the query-translation approach.
Tento článek popisuje účast týmu CUNI-MTIR v soutěži COVID-19 MLIA @ Eval Task 2.
[ "Shadi Saleh", "Hadi Abdi Khojasteh", "Hashem Sellat", "Pavel Pecina" ]
https://www.semanticscholar.org/paper/80a0c9a8ec5b23bc0820533809ff460ef9361fe5/
null
null
null
null
null
null
null
null
null
null
null
null
cs
2021
Artificial creative intelligence
Umělá kreativní inteligence
Can AI create works of art? Music, painting, literature, film, theatre? Or something completely new? Can a neural net stir emotions? Engage the machines in the creation along with the THEaiTRE challenge guarantor.
Dokáže AI tvořit umělecká díla? Hudbu, malířství, literaturu, film, divadlo? Nebo něco úplně nového? Dokáže neuronová síť vzbudit emoce? Zapojte stroje do tvorby spolu s garantem výzvy THEaiTRE.
[ "Tomáš Studeník", "Rudolf Rosa", "Tomáš Musil", "David Košťák", "Martina Kinská" ]
https://www.semanticscholar.org/paper/25a25e355e1c44a228ddda0512b5d716b686f787/
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
Representations of Meaning in Neural Networks for NLP: a Thesis Proposal
Reprezentace významu v neuronových sítích pro NLP
Neural networks are the state-of-the-art method of machine learning for many problems in NLP. Their success in machine translation and other NLP tasks is phenomenal, but their interpretability is challenging. We want to find out how neural networks represent meaning. In order to do this, we propose to examine the distribution of meaning in the vector space representation of words in neural networks trained for NLP tasks. Furthermore, we propose to consider various theories of meaning in the philosophy of language and to find a methodology that would enable us to connect these areas.
Neuronové sítě jsou nejmodernější metodou strojového učení pro mnoho problémů v NLP. Jejich úspěch ve strojovém překladu a dalších úlohách NLP je fenomenální, ale jejich interpretovatelnost je náročná. Chceme zjistit, jak neuronové sítě reprezentují význam. Za tímto účelem navrhujeme prozkoumat rozložení významu ve vektorovém prostoru reprezentace slov v neuronových sítích trénovaných pro úlohy NLP. Dále navrhujeme zvážit různé teorie významu ve filosofii jazyka a najít metodologii, která by nám umožnila tyto oblasti propojit.
[ "Tomáš Musil" ]
https://www.semanticscholar.org/paper/cd8e65d8feb00c54aa0c9e045fdfe51b65ac9bbe/
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
Shades of BLEU, Flavours of Success: The Case of MultiWOZ
Odstíny BLEU, příchutě Success: případ MultiWOZ
The MultiWOZ dataset (Budzianowski et al.,2018) is frequently used for benchmarking context-to-response abilities of task-oriented dialogue systems. In this work, we identify inconsistencies in data preprocessing and reporting of three corpus-based metrics used on this dataset, i.e., BLEU score and Inform & Success rates. We point out a few problems of the MultiWOZ benchmark such as unsatisfactory preprocessing, insufficient or under-specified evaluation metrics, or rigid database. We re-evaluate 7 end-to-end and 6 policy optimization models in as-fair-as-possible setups, and we show that their reported scores cannot be directly compared. To facilitate comparison of future systems, we release our stand-alone standardized evaluation scripts. We also give basic recommendations for corpus-based benchmarking in future works.
Dataset MultiWOZ (Budzianowski et al.,2018) je často užíván na poměřování schopností generovat odpověď z kontextu v případě dialogových systému zaměřených na úkoly. V této práci identifikujeme nekonzistence v předzpracování dat a reportování tří metrik založených na evaluačním korpusu, tj., BLEU skóre, míry Inform a míry Success, v kontextu tohoto datasetu. Poukazujeme na několik problémů benchmarku MultiWOZ jako je neuspokojivé předzpracování dat, nedostatečné nebo nedostatečně specifikované evaluační metriky, nebo neohebná databáze. Ve spravedlivých podmínkách jsme znovu vyhodnotili 7 end-to-end a 6 policy optimization modelů a ukázali jsme, že jejich původně reportovaná skóre nemohou být přímo srovnávána. Abychom ulehčili porovnávání budoucích systémů, zveřejňujeme naše soběstačné standardizované evaluační skripty. Rovněž dáváme základní doporučení pro budoucí vyhodnocování založená na evaluačním korpusu.
[ "Tomáš Nekvinda", "Ondřej Dušek" ]
https://www.semanticscholar.org/paper/d586ece0db87c7ae8dfcb87d18f6a26d0695ec09/
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
Overview of the Second Workshop on Scholarly Document Processing
Přehled druhého semináře o zpracovávání dokumentů v oblasti vědy
With the ever-increasing pace of research and high volume of scholarly communication, scholars face a daunting task. Not only must they keep up with the growing literature in their own and related fields, scholars increasingly also need to rebut pseudo-science and disinformation. These needs have motivated an increasing focus on computational methods for enhancing search, summarization, and analysis of scholarly documents. However, the various strands of research on scholarly document processing remain fragmented. To reach out to the broader NLP and AI/ML community, pool distributed efforts in this area, and enable shared access to published research, we held the 2nd Workshop on Scholarly Document Processing (SDP) at NAACL 2021 as a virtual event (https://sdproc.org/2021/). The SDP workshop consisted of a research track, three invited talks and three Shared Tasks (LongSumm 2021, SCIVER and 3C). The program was geared towards NLP, information retrieval, and data mining for scholarly documents, with an emphasis on identifying and providing solutions to open challenges.
Se stále rostoucím tempem výzkumu a velký objem vědecké komunikace, stipendisty čeká nelehký úkol. Nejen musí udržují krok s rostoucí literaturou v svých vlastních a souvisejících oborů, musí vědci stále častěji také vyvracet pseudovědu a dezinformace. Tyto potřeby motivovaly rostoucí zaměření na výpočetní metody pro zlepšení vyhledávání, sumarizace a analýzy odborných dokumentů. Nicméně různé oblasti výzkumu vědeckého zpracování dokumentů zůstávají roztříštěné. K dosažení širší komunitě NLP a AI/ML, společné úsilí v této oblasti a umožnit sdílený přístup k publikovanému výzkumu, my uspořádal 2. seminář o zpracovávání dokumentů (SDP) v rámci programu NAACL 2021 jako virtuální událost (https://sdproc.org/2021/). The SDP workshop sestával z výzkumné dráhy, tři pozvané rozhovory a tři sdílené úkoly (LongSumm 2021, SCIVER a 3C). Program byl zaměřen na NLP, získávání informací a dolování dat pro vědecké dokumenty s důrazem na identifikaci a řešení otevřených výzev.
[ "Iz Beltagy", "Arman Cohan", "Guy Feigenblat", "Dayne Freitag", "Tirthankar Ghosal", "Keith Hall", "Drahomira Hermannova", "Petr Knoth", "Kyle Lo", "Philipp Mayr", "Robert Patton", "Michal Shmueli-Scheuer", "Anita De-Waard", "Kuansan Wang", "Lucy Lu Wang" ]
https://www.semanticscholar.org/paper/f25a978da0f34c98db85c8c6f1613570b9c8a841/
null
null
null
null
null
null
null
null
null
null
null
null
cs
2021
When a robot writes a play, it has an unwanted sense of humor and thinks about sex a lot
Když robot píše hru, má nechtěný smysl pro humor a dost myslí na sex
The Švandovo Theatre is about to premiere the first Czech play, the dialogue of which was "written" by an artificial intelligence. "AI: When a robot writes a play" was created in collaboration with the Faculty of Mathematics and Physics of Charles University and commemorates a century since the release of the play R.U.R. by Karel Čapek, in which the word "robot" was first mentioned.
Švandovo divadlo chystá premiéru první české hry, jejíž dialogy „napsala“ umělá inteligence. Inscenace AI: Když robot píše hru vznikla ve spolupráci s Matematicko-fyzikální fakultou Univerzity Karlovy a připomíná sto let od uvedení hry R.U.R. od Karla Čapka, v níž slovo „robot“ poprvé zaznělo.
[ "Martin Tyburec", "Rudolf Rosa", "David Košťák", "Ondřej Dušek", "Daniel Hrbek" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
ELITR Coordination Experience
Zkušenosti s koordinací projektu ELITR
Describing our experience from the coordination of the EU project ELITR for grant applicants.
Seznámení uchazečů o různé projekty našimi zkušenostmi s koordinací projektu ELITR.
[ "Ondřej Bojar" ]
https://www.semanticscholar.org/paper/c38b7c756c99be0e489b6ecc9810422e7eddd9ca/
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
ELITR Demo at LCT Summer School
Ukázka překladače ELITR na letní škole LCT
I presented the research project ELITR and the state of the art in machine translation and speech translation.
Představil jsem výzkumný projekt ELITR a současný stav poznání v oblasti strojového překladu a překladu mluvené řeči.
[ "Ondřej Bojar" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
Multi-Lingual MT Research at ÚFAL, Charles University
Výzkum mnohojazyčného překladu na ÚFALu, Univerzita Karlova
A brief description of two applied projects, ELITR and Bergamot, and one basic research project NEUREM3, run at UFAL
Představení dvou aktuálních aplikovaných projektů ELITR a Bergamot a projektu základního výzkumu NEUREM3, realizovaných na ÚFALu.
[ "Ondřej Bojar" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
cs
2021
Computational linguistics: machine translation and interpreting
Počítačová lingvistika: strojové překlady a tlumočení
I presented the area of computational linguitics with a focus on the paradigm change brought by neural networks, in articular in the application area of machine translation and interpreting.
Představil jsem obor počítačové lingvistiky s důrazem na změnu paradigmatu zavedením neuronových sítí, zejména v oblasti strojového překladu a tlumočení.
[ "Ondřej Bojar" ]
https://www.semanticscholar.org/paper/19dcc665dee21a8148f9bd07066332783d212126/
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
ELITR Multilingual Live Subtitling: Demo and Strategy
Simultánní mnohojazyčné titulkování projektu ELITR: Demo a strategie
This paper presents an automatic speech translation system aimed at live subtitling of conference presentations. We describe the overall architecture and key processing components. More importantly, we explain our strategy for building a complex system for end-users from numerous individual components, each of which has been tested only in laboratory conditions. The system is a working prototype that is routinely tested in recognizing English, Czech, and German speech and presenting it translated simultaneously into 42 target languages.
542 / 5000 Translation results Tento článek představuje systém automatického překladu řeči zaměřený na živé titulkování konferenčních prezentací. Popisujeme celkovou architekturu a klíčové komponenty zpracování. Důležitější je, že vysvětlujeme naši strategii budování komplexního systému pro koncové uživatele z mnoha jednotlivých komponent, z nichž každá byla testována pouze v laboratorních podmínkách. Systém je funkčním prototypem, který je rutinně testován v rozpoznávání anglické, české a německé řeči a současně je prezentován přeložený do 42 cílových jazyků.
[ "Ondřej Bojar", "Dominik Macháček", "Sangeet Sagar", "Otakar Smrž", "Jonáš Kratochvíl", "Peter Polák", "Ebrahim Ansari", "Mohammad Mahmoudi", "Rishu Kumar", "Dario Franceschini", "Chiara Canton", "Ivan Simonini", "Thai-Son Nguyen", "Felix Schneider", "Sebastian Stüker", "Alex Waibel", "Barry Haddow", "Rico Sennrich", "Philip Williams" ]
https://www.semanticscholar.org/paper/6e45bac68d46d9832721add3c67f74e8ee070889/
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
Operating a Complex SLT System with Speakers and Human Interpreters
Provoz komplexního systému pro překlad mluvené řeči s řečníky a tlumočníky
The paper describes the practical experience from testing our complex system translating from the speech of speakers and interpreters.
Příspěvek popisuje praktické zkušenosti z testování systému pro simultánní překlad mluvené řeči z projevu řečníků a tlumočníků.
[ "Ondřej Bojar", "Vojtěch Srdečný", "Rishu Kumar", "Otakar Smrž", "Felix Schneider", "Barry Haddow", "Phil Williams", "Chiara Canton" ]
https://www.semanticscholar.org/paper/c63834f38acdffafff85eb81e84afb2910a3fb20/
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
FAUST cs-en 0.5
FAUST cs-en 0.5
This machine translation test set contains 2223 Czech sentences collected within the FAUST project (https://ufal.mff.cuni.cz/grants/faust, http://hdl.handle.net/11234/1-3308). Each original (noisy) sentence was normalized (clean1 and clean2) and translated to English independently by two translators.
Tato testovací sada strojového překladu obsahuje 2223 českých vět shromážděných v rámci projektu FAUST (https://ufal.mff.cuni.cz/grants/faust, http://hdl.handle.net/11234/1-3308). Každá původní věta obsahující šum byla normalizována (clean1 a clean2) a nezávisle na sobě přeložena do angličtiny dvěma překladateli.
[ "Jan Hajič", "David Mareček", "Eva Fučíková", "Silvie Cinková", "Jan Štěpánek", "Marie Mikulová", "Martin Popel" ]
https://www.semanticscholar.org/paper/bd528af41d152c1ac9e683ca5908983261cb5b98/
null
null
null
null
null
null
null
null
null
null
null
null
cs
2021
On selected types of deverbative adjectives in Czech (in comparison with Russian)
K vybraným typům deverbativních adjektiv v češtině (ve srovnání s ruštinou)
Selected Czech adjectives derived from the verbs by the suffix -lý are analyzed and compared with the other types of deverbal adjectives and their mutual interchanges are presented in comparison with their independent meanings. The Russian system of deverbal adjectives is briefly demonstrated as to their position different from Czech system.
Ve stati se analyzují na základě korpusových dat vybraná adjektiva tvořená od sloves příponou -lý a porovnávají se s dalšími typy slovesných adjektiv adjektiv z hlediska vzájemné zaměnitelnosti a významových rozdílů.
[ "Jarmila Panevová" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
cs
2021
A survey into Czech word formation: verbs formed from verbs or verbs formed from nouns?
Sonda do české slovotvorby: Substantiva utvořená od sloves nebo slovesa utvořená od substantiv?
The study deals with a topic in Czech word formation, namely with the determination of the direction of formation in pairs of a suffixless noun and a related verb (e.g. skok "jump" - skočit "to jump", sůl "salt" - solit "to salt"). While corpus frequency is an unreliable criterion due to formal-meaning asymmetries and meaning shifts, alternations in the prefix and root morphemes of the studied nouns and verbs, as well as the way of formation of the aspectual counterparts, are useful clues - the latter criterion can be applied to both domestic and foreign verbs, including neologisms.
Studie se zabývá tématem z české slovotvorby, konkrétně stanovenám směru tvoření ve dvojicích bezpříponového substantiva a příbuzného slovesa (např. skok - skočit, sůl - solit). Hledá kritéria, která by směr tvoření v těchto dvojicích pomohla určit. Zatímco korpusová frekvence je z důvodu formálně-významových asymetrií a významových posunů kritériem nespolehlivým, užitečným vodítkem jsou alternace v předponových i kořenných morfémech zkoumaných substantiv a sloves a také způsob tvoření vidových protějšků – toto kritérium se dá aplikovat na slovesa domácí i cizí, včetně neologismů.
[ "Jarmila Panevová", "Magda Ševčíková" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
cs
2021
A Few Remarks on Word Order and Topic-Focus Articulation in the ligtht of an English-Czech parallel corpus
Několik poznámek ke slovosledu a aktuálnímu členění ve světle anglicko-českého paralelního korpusu
In this paper, we study the relationship between word order and topic-focus articulation in comparison of English and Czech, namely in two areas: the position of proverbial determinations of time and place in the focal part of a sentence and the semantic range of particles called focalizers. We used the material of an annotated parallel English-Czech corpus (PCEDT).
V příspěvku sledujeme vztah slovosledu a aktuálního členění ve srovnání angličtiny s češtinou, a to ve dvou oblastech: umístění příslovečných určení času a místa v ohniskové části věty a sémantického dosahu částic zvaných fokalizátory. Materiálem nám byl anotovaný paralelní korpus anglicko–český (PCEDT).
[ "Eva Hajičová", "Jiří Mírovský", "Barbora Štěpánková" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
cs
2021
Open Science and FAIR aspects in the LINDAT/CLARIAH-CZ repository
Aspekty Open Science a FAIR v repozitáři LINDAT/CLARIAH-CZ
Introducing the functionality of the LINDAT/CLARIAH-CZ repository, that is, mainly CLARIN/Dspace, which supports the principles of Open Science, primarily the so-called FAIR principles.
Představení funkcionality repozitáře LINDAT/CLARIAH-CZ, tedy hlavně CLARIN/Dspace, která podporuje principy Open Science, v první řadě tzv. FAIR principy.
[ "Pavel Straňák" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
Data citations at LINDAT/CLARIAH-CZ
Citování dat v LINDAT/CLARIAH-CZ
Why and how the LINDAT data repository integrates services supporting and using direct data citation.
Proč a jakým způsobem integruje datový repozitář LINDATu služby podporující a využívající přímé citování dat.
[ "Pavel Straňák" ]
https://www.semanticscholar.org/paper/b8be943b6b65b6f54c776175354a39a9e47c076b/
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
LINDAT/CLARIAH-cz Repository
Repozitář LINDAT/CLARIAH-cz
Information about the central repository of LINDAT/CLARIAH-cz, especially in terms of FAIR aspects
Informace o centrálním repozitáři LINDAT/CLARIAH-cz, obzvláště z hlediska FAIR aspektů
[ "Pavel Straňák" ]
https://www.semanticscholar.org/paper/b95995b020d9ac4ab498ec9edc4969cdaf7f5837/
null
null
null
null
null
null
null
null
null
null
null
null
cs
2021
LINDAT/CLARIAH-CZ: Subject-specific repository not only for language data
LINDAT/CLARIAH-CZ: Oborový repozitář nejen pro jazyková data
The benefits of data publishing, especially with high-quality certrified repositories well integrated with other infrastructure.
Přinosy zveřejňování dat, obzvláště pomocí kvalitních certrifikovaných repozitářů dobře integrovaných s další infrastrukturou.
[ "Pavel Straňák" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
cs
2021
LINDAT/CLARIAH-CZ: field-specific repository for not only language data
LINDAT/CLARIAH-CZ: oborový repozitář nejen pro jazyková data
We will present a repository for language data and other data primarily from the humanities. The repository is operated by the large research infrastructure (LRI) LINDAT/CLARIAH-CZ and allows for easy, safe and long-term preservation, sharing and direct quoting of scientific data. We have expanded the repository by our own development to include a module that allows data to assign any license and sign this, if necessary, online. We will briefly explain why such a solution is necessary. We also address FAIR aspects of the repository, and why it is far better to store our research data with us than to share it with our own website and other ad hoc methods. The repository also includes a solution for selecting the most open possible license for the data: Public License Selector. The user answers questions when saving data, navigating it through a maze of public licenses to the most appropriate option for the data or software they want to share. The two connected solutions we will present are open source, developed publicly under LRI LINDAT/CLARIAH-CZ (https://github.com/ufal/clarin-dspace, https://github.com/ufal/public-license-selector) and used in a variety of installations across Europe.
Představíme repozitář pro jazyková data a jiná data primárně z humanitních oborů. Repozitář je provozován velkou výzkumnou infrastrukturou (VVI) LINDAT/CLARIAH-CZ a umožňuje snadné, bezpečné a dlouhodobé uchování, sdílení a přímé citování vědeckých dat. Repozitář jsme vlastním vývojem rozšířili o modul, který umožňuje datům přiřadit libovolnou licenci a tuto případně i online podepsat. Stručně vysvětlíme, proč je takové řešení nutné. Věnujeme se také FAIR aspektům repozitáře a tomu, proč je mnohem lepší svá výzkumná data uložit k nám, než je sdílet pomocí své vlastní webové stránky a jinými ad hoc metodami. Součástí repozitáře je také řešení pro výběr co nejotevřenější možné licence pro daná data: Public License Selector. uživatel při ukládání dat odpovídá na otázky, které jej navigují bludištěm veřejných licencí k nejvhodnější možnosti pro daná data nebo software, jež chce sdílet. Obě propojená řešení, která představíme, jsou open source, vyvíjená veřejně v rámci VVI LINDAT/CLARIAH-CZ (https://github.com/ufal/clarin-dspace, https://github.com/ufal/public-license-selector) a jsou používána v řadě instalací v celé Evropě.
[ "Pavel Straňák" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
cs
2021
Open Science support in LINDAT/CLARIAH-CZ repository
Podpora Open Science v repozitáři LINDAT/CLARIAH-CZ
Introducing the functionality of the LINDAT/CLARIAH-CZ repository, that is, mainly CLARIN/Dspace, which supports the principles of Open Science, primarily the so-called FAIR principles.
Představení funkcionality repozitáře LINDAT/CLARIAH-CZ, tedy hlavně CLARIN/Dspace, která podporuje principy Open Science, v první řadě tzv. FAIR principy.
[ "Pavel Straňák" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
Annotation guidelines for German verbal synonyms included in SynSemClass Lexicon
Manuál pro anotaci německých synonymních sloves ve slovníku SynSemClass
This report presents a guideline for including German synonyms into the multilingual SynSemClass lexicon.
Technická zprava shrnuje pravidla pro zařazení německých synonym do synonymního slovníku SynSemClass.
[ "Zdeňka Urešová", "Eva Fučíková", "Jan Hajič", "Karolina Zaczynska" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
SynSemClass 3.5
SynSemClass 3.5
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. SynSemClass3.5 is a multilingual event-type ontology based on classes of synonymous verb senses, complemented with semantic roles and links to existing semantic lexicons. Apart of 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 - by 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 3.5 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 3.5 je vícejazyčná ontologie typu události založená na třídách synonymních významů sloves, doplněná sémantickými rolemi a odkazy na existující sémantické lexikony. 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", "Eva Fučíková", "Eva Hajičová", "Jan Hajič", "Georg Rehm", "Karolina Zaczynska" ]
https://www.semanticscholar.org/paper/6cb63773f1924add840f59357e988cf2481672a6/
null
null
null
null
null
null
null
null
null
null
null
null
cs
2021
PDT-Vallex: Czech Valency lexicon linked to treebanks 4.0
PDT-Vallex: Valenční slovník češtiny propojený s korpusy 4.0
PDT-Vallex is a Czech valence lexicon linked to real texts in several Czech corpora and is also part of PDT-C 1.0. PDT-Vallex 4.0 contains 14528 valence frames for 8498 verbs that occurred in PDT-C 1.0. PDT-Vallex 4.0 is an expanded version of the original PDT-Vallex, which contained 7121 verb entries with 11933 frames.
PDT-Vallex je český valenční lexikon propojený s reálnými texty v několika českých korpusech a je také součástí PDT-C 1.0. PDT-Vallex 4.0 obsahuje 14528 valenčních rámců pro 8498 sloves, která se vyskytla v PDT-C 1.0. PDT-Vallex 4.0 je rozšířenou verzí původního PDT-Vallexu, který obsahoval 7121 slovesných záznamů s 11933 rámci.
[ "Zdeňka Urešová", "Alevtina Bémová", "Eva Fučíková", "Jan Hajič", "Veronika Kolářová", "Marie Mikulová", "Petr Pajas", "Jarmila Panevová", "Jan Štěpánek" ]
https://www.semanticscholar.org/paper/39b9394be0cf60a671d201a8074ccbe473ee93b5/
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
Systemic and Non-systemic Valency Behavior of Czech Deverbal Adjectives
Systémové a nesystémové valenční chování českých deverbálních adjektiv
We present results of an automatic comparison of valency frames of interlinked adjectival and verbal lexical units based on the valency lexicons NomVallex and VALLEX. We distinguish nine derivational types of deverbal adjectives and examine whether they tend to display systemic valency behavior, or rather the non-systemic one. The non-systemic valency behavior includes changes in the number of valency complementations and, more dominantly, non-systemic forms of actants, especially a prepositional group.
V článku prezentujeme výsledky automatického srovnání valenčních rámců vzájemně propojených adjektivních a slovesných lexikálních jednotek, obsažených ve valenčních slovnících NomVallex a VALLEX. Rozlišujeme devět derivačních typů deverbálních adjektiv a zkoumáme, zda vykazují systémové valenční chování, nebo spíše nesystémové valenční chování. K projevům nesystémového valenčního chování patří změny v počtu valenčních doplnění a především nesystémové formy aktantů, zvláště nesystémová předložková skupina.
[ "Veronika Kolářová", "Anna Vernerová", "Jana Klímová" ]
https://www.semanticscholar.org/paper/2621f3f3a60a692afd4189c4211522943d3a5e7a/
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
Extending Coverage of a Lexicon of Discourse Connectives Using Annotation Projection
Rozšíření pokrytí Slovníku českých diskurzních konektorů pomocí anotační projekce
We present a method for extending coverage of the Lexicon of Czech Discourse Connectives – CzeDLex – using annotation projection. We take advantage of two language resources: (i) the Penn Discourse Treebank 3.0 as a source of manually annotated discourse relations in English, and (ii) the Prague Czech–English Dependency Treebank 2.0 as a translation of the English texts to Czech and a link between tokens on the two language sides. Although CzeDLex was originally extracted from a large Czech corpus, the presented method resulted in an addition of a number of new connectives and new types of usages (discourse types) for already present entries in the lexicon.
Představujeme metodu pro rozšíření pokrytí Slovníku českých diskurzních konektorů (CzeDLex) pomocí anotační projekce. Využíváme dva jazykové zdroje: (i) Penn Discourse Treebank 3.0 jako zdroj ručně anotovaných diskurzních vztahů v angličtině a (ii) Pražský česko -anglický závislostí korpus 2.0 jako překlad anglických textů do češtiny a propojení mezi tokeny v obou jazycích. Přestože byl CzeDLex původně extrahován z velkého českého korpusu, vedla prezentovaná metoda k přidání řady nových konektorů a nových užití (diskurzních typů) pro již přítomné položky ve slovníku.
[ "Jiří Mírovský", "Pavlína Synková", "Lucie Poláková" ]
https://www.semanticscholar.org/paper/c757fd22ca43674c2bf3b9d8a4a1ffea0dbbcb23/
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
CzeDLex 1.0
CzeDLex 1.0
CzeDLex 1.0 is the first production version of a lexicon of Czech discourse connectives, following three previous development versions. The lexicon contains connectives partially automatically extracted from the Prague Discourse Treebank 2.0 (PDiT 2.0) and other resources. All entries in the lexicon have been manually checked, translated to English and supplemented with additional linguistic information.
CzeDLex 1.0 je první produkční verze slovníku českých diskurzních konektorů, navazující na 3 předchozí verze vývojové. Slovník obsahuje konektory částečně automaticky extrahované z Pražského diskurzního korpusu 2.0 (PDiT 2.0) a z dalších zdrojů. Všechna slovníková hesla byla ručně zkontrolována, přeložena do angličtiny a doplněna dalšími lingvistickými informacemi.
[ "Jiří Mírovský", "Pavlína Synková", "Lucie Poláková", "Věra Kloudová", "Magdaléna Rysová" ]
https://www.semanticscholar.org/paper/1ae29c65f7cc846e639859635687d0276be2ea00/
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
Sense Prediction for Explicit Discourse Relations with BERT
Predikce významu explicitních diskurzních vztahů systémem BERT
We present experiments in predicting a discourse sense for explicit inter-sentential discourse relations in Czech and English, using embedding and deep learning (fine-tuned BERT) to predict the senses, and annotation projection from English to Czech to increase the size of training data for Czech.
Předkládáme experimenty zaměřené na predikci významu explicitních mezivětných diskurzních vztahů v češtině a angličtině, s využitím hlubokého učení (BERT) pro predikci významů a anotační projekce z angličtiny do češtiny pro zvětšení množství trénovacích dat pro češtinu.
[ "Jiří Mírovský", "Lucie Poláková" ]
https://www.semanticscholar.org/paper/e607f28c04185577b3a90ccb66bd447b9a856733/
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
Date and Time in Universal Dependencies
Datum a čas v Universal Dependencies
We attempt to shed some light on the various ways how languages specify date and time, and on the options we have when trying to annotate them uniformly across Universal Dependencies. Examples from several language families are discussed, and their annotation is proposed. Our hope is to eventually make this (or similar) proposal an integral part of the UD annotation guidelines, which would help improve consistency of he UD treebanks. The current annotations are far from consistent, as can be seen from the survey we provide in appendices to this paper.
Pokoušíme se osvětlit rozličné způsoby, kterými jazyky udávají datum a čas, a možnosti, které máme, když se příslušné konstrukce snažíme jednotně zachytit ve formalismu Universal Dependencies. Probíráme příklady z několika jazykových rodin a navrhujeme jejich anotaci. Doufáme, že tento (nebo podobný) návrh by se mohl v budoucnosti stát součástí anotačních pravidel UD, což by přispělo k větší konzistenci treebanků UD. Současné anotace mají ke konzistenci daleko, jak přehledně ukazujeme v dodatcích k této studii.
[ "Daniel Zeman" ]
https://www.semanticscholar.org/paper/9653fb135ce68c02b2da572bf10a09e5012a3bf1/
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
Enhanced Universal Dependencies: The Current State and Outlook
Enhanced Universal Dependencies: Aktuální stav a výhled
Universal Dependencies (UD) is a multilingual collection of corpora featuring morphological and syntactic annotation in a unified style. We discuss an optional layer of deep-syntactic annotation in UD, called Enhanced Universal Dependencies. We survey the existing enhanced representation as of release 2.8 and consider two possible future expansions: semi-automatic addition of existing enhancement types to new languages, and addition of new enhancement types.
Universal Dependencies (UD, univerzální závislosti) je mnohojazyčná kolekce korpusů, morfologicky a syntakticky anotovaných v jednotném stylu. Představujeme volitelnou rovinu hloubkově-syntaktické anotace v UD, zvanou Enhanced Universal Dependencies (rozšířené univerzální závislosti). Podáváme přehled rozšířených anotací ve verzi 2.8 a zvažujeme dvě možná budoucí pokračování: poloautomatické přidávání známých typů rozšíření do nových jazyků a přidávání nových typů rozšíření.
[ "Daniel Zeman" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
Universal Dependencies: Comparing Languages in Space and Time
Universal Dependencies: Srovnávání jazyků v prostoru a čase
I present Universal Dependencies, a worldwide community project to design morphological and syntactic annotation guidelines applicable to all the languages of the world. I discuss automatic parsing models and their use in digital humanities: linguistics, language teaching, documentation of endangered languages, linguistic typology and language change.
Představuji Universal Dependencies, mezinárodní komunitní projekt, v jehož rámci vznikla morfologická a syntaktická anotační pravidla aplikovatelná na všechny přirozené jazyky světa. Probírám modely pro automatickou syntaktickou analýzu a jejich využití v digitálních humanitních studiích: v lingivistice, výuce jazyků, dokumentaci ohrožených jazyků, jazykové typologii a historickém vývoji jazyků.
[ "Daniel Zeman" ]
https://www.semanticscholar.org/paper/d83e97b195ac104cbe0817870e382d71574ae0ef/
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
Universal Description of Morphology and Syntax of Natural Languages
Univerzální popis morfologie a syntaxe přirozených jazyků
I will present Universal Dependencies, 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. From the engineering perspective, UD treebanks serve as training data for automatic parsers that can be subsequently used to analyze previously unseen text. The parsed output is an intermediate representation between the input text and its underlying meaning. It is helpful in foreign language learning as well as for automatic extraction of semantic relations (answers to “who did what to whom”). I will thus discuss these semantic aspects in the last part of my talk; in particular, I will look at extensions of UD that have been proposed and that focus more on deep-syntactic and semantic relations expressed in natural language.
Představím Universal Dependencies, mezinárodní komunitní projekt a kolekci morfosyntakticky anotovaných datových sad (treebanků) pro více než 100 jazyků. Tato kolekce je neocenitelným zdrojem pro různé lingvistické studie od gramatických konstrukcí v jednom jazyce k jazykové typologii, dokumentaci ohrožených jazyků a historickému vývoji jazyka.
[ "Daniel Zeman" ]
https://www.semanticscholar.org/paper/34d9e05090c2b0cf782d2f90e27f41603a049772/
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
Deep Universal Dependencies 2.8
Deep Universal Dependencies 2.8
Deep Universal Dependencies is a collection of treebanks derived semi-automatically from Universal Dependencies. It contains additional deep-syntactic and semantic annotations. Version of Deep UD corresponds to the version of UD it is based on. Note however that some UD treebanks have been omitted from Deep UD.
Deep Universal Dependencies (hluboké univerzální závislosti) je sbírka treebanků poloautomaticky odvozených z Universal Dependencies. Obsahuje přídavné hloubkově-syntaktické a sémantické anotace. Verze Deep UD odpovídá verzi UD, na které je založena. Mějte však na paměti, že některé treebanky UD nebyly do Deep UD zahrnuty.
[ "Daniel Zeman", "Kira Droganova" ]
https://www.semanticscholar.org/paper/a2d25cf45f034108628a044b5470ed358164fc02/
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
IWPT 2021 Shared Task Data and System Outputs
Data a systémové výstupy soutěže IWPT 2021
This package contains data used in the IWPT 2021 shared task. It contains training, development and test (evaluation) datasets. The data is based on a subset of Universal Dependencies release 2.7 (http://hdl.handle.net/11234/1-3424) but some treebanks contain additional enhanced annotations. Moreover, not all of these additions became part of Universal Dependencies release 2.8 (http://hdl.handle.net/11234/1-3687), which makes the shared task data unique and worth a separate release to enable later comparison with new parsing algorithms. The package also contains a number of Perl and Python scripts that have been used to process the data during preparation and during the shared task. Finally, the package includes the official primary submission of each team participating in the shared task.
Tento balíček obsahuje data použitá v soutěži IWPT 2021. Obsahuje trénovací, vývojová a testovací (vyhodnocovací) datové množiny. Data jsou založena na podmnožině vydání 2.7 Universal Dependencies (http://hdl.handle.net/11234/1-3424), ale některé treebanky obsahují další obohacené anotace nad rámec UD 2.7.
[ "Daniel Zeman", "Gosse Bouma", "Djamé Seddah" ]
https://www.semanticscholar.org/paper/965e2c78db7885960e65d001d65c314603a69b1c/
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
Universal Dependencies 2.8
Universal Dependencies 2.8
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 fourteenth release of UD Treebanks, Version 2.8.
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 čtrnácté vydání treebanků UD, verze 2.8.
[ "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", "Lene Antonsen", "Katya Aplonova", "Angelina Aquino", "Carolina Aragon", "Maria Jesus Aranzabe", "Bilge Nas Arıcan", "Þórunn Arnardóttir", "Gashaw Arutie", "Jessica Naraiswari Arwidarasti", "Masayuki Asahara", "Deniz Baran Aslan", "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", "Victoria Basmov", "Colin Batchelor", "John Bauer", "Seyyit Talha Bedir", "Kepa Bengoetxea", "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", "Alessandra T. Cignarella", "Silvie Cinková", "Aurélie Collomb", "Çağrı Çöltekin", "Miriam Connor", "Marine Courtin", "Mihaela Cristescu", "Philemon Daniel", "Elizabeth Davidson", "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", "Marília Fernanda", "Hector Fernandez Alcalde", "Jennifer Foster", "Cláudia Freitas", "Kazunori Fujita", "Katarína Gajdošová", "Daniel Galbraith", "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", "Sam Hardwick", "Kim Harris", "Dag Haug", "Johannes Heinecke", "Oliver Hellwig", "Felix Hennig", "Barbora Hladká", "Jaroslava Hlaváčová", "Florinel Hociung", "Petter Hohle", "Eva Huber", "Jena Hwang", "Takumi Ikeda", "Anton Karl Ingason", "Radu Ion", "Elena Irimia", "Ọlájídé Ishola", "Kaoru Ito", "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", "Boris Katz", "Tolga Kayadelen", "Jessica Kenney", "Václava Kettnerová", "Jesse Kirchner", "Elena Klementieva", "Arne Köhn", "Abdullatif Köksal", "Kamil Kopacewicz", "Timo Korkiakangas", "Natalia Kotsyba", "Jolanta Kovalevskaitė", "Simon Krek", "Parameswari Krishnamurthy", "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", "Andry Luthfi", "Mikko Luukko", "Olga Lyashevskaya", "Teresa Lynn", "Vivien Macketanz", "Aibek Makazhanov", "Michael Mandl", "Christopher Manning", "Ruli Manurung", "Büşra Marşan", "Cătălina Mărănduc", "David Mareček", "Katrin Marheinecke", "Héctor Martínez Alonso", "André Martins", "Jan Mašek", "Hiroshi Matsuda", "Yuji Matsumoto", "Alessandro Mazzei", "Ryan McDonald", "Sarah McGuinness", "Gustavo Mendonça", "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", "Petya Osenova", "Robert Östling", "Lilja Øvrelid", "Şaziye Betül Özateş", "Merve Özçelik", "Arzucan Özgür", "Balkız Öztürk Başaran", "Hyunji Hayley Park", "Niko Partanen", "Elena Pascual", "Marco Passarotti", "Agnieszka Patejuk", "Guilherme Paulino-Passos", "Angelika Peljak-Łapińska", "Siyao Peng", "Cenel-Augusto Perez", "Natalia Perkova", "Guy Perrier", "Slav Petrov", "Daria Petrova", "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", "Taraka Rama", "Loganathan Ramasamy", "Carlos Ramisch", "Fam Rashel", "Mohammad Sadegh Rasooli", "Vinit Ravishankar", "Livy Real", "Petru Rebeja", "Siva Reddy", "Georg Rehm", "Ivan Riabov", "Michael Rießler", "Erika Rimkutė", "Larissa Rinaldi", "Laura Rituma", "Luisa Rocha", "Eiríkur Rögnvaldsson", "Mykhailo Romanenko", "Rudolf Rosa", "Valentin Roșca", "Davide Rovati", "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", "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", "Carolyn Spadine", "Rachele Sprugnoli", "Steinþór Steingrímsson", "Antonio Stella", "Milan Straka", "Emmett Strickland", "Jana Strnadová", "Alane Suhr", "Yogi Lesmana Sulestio", "Umut Sulubacak", "Shingo Suzuki", "Zsolt Szántó", "Dima Taji", "Yuta Takahashi", "Fabio Tamburini", "Mary Ann C. Tan", "Takaaki Tanaka", "Samson Tella", "Isabelle Tellier", "Marinella Testori", "Guillaume Thomas", "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", "Sowmya Vajjala", "Rob van der Goot", "Martine Vanhove", "Daniel van Niekerk", "Gertjan van Noord", "Viktor Varga", "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", "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", "Zdeněk Žabokrtský", "Shorouq Zahra", "Amir Zeldes", "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
2021
Universal Dependencies 2.9
Universal Dependencies 2.9
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 fifteenth release of UD Treebanks, Version 2.9.
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 patnácté vydání treebanků UD, verze 2.9.
[ "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", "Lene Antonsen", "Katya Aplonova", "Angelina Aquino", "Carolina Aragon", "Maria Jesus Aranzabe", "Bilge Nas Arıcan", "Þórunn Arnardóttir", "Gashaw Arutie", "Jessica Naraiswari Arwidarasti", "Masayuki Asahara", "Deniz Baran Aslan", "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", "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", "Marine Courtin", "Mihaela Cristescu", "Philemon Daniel", "Elizabeth Davidson", "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", "Jannatul Ferdaousi", "Marília Fernanda", "Hector Fernandez Alcalde", "Jennifer Foster", "Cláudia Freitas", "Kazunori Fujita", "Katarína Gajdošová", "Daniel Galbraith", "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", "Sam Hardwick", "Kim Harris", "Dag Haug", "Johannes Heinecke", "Oliver Hellwig", "Felix Hennig", "Barbora Hladká", "Jaroslava Hlaváčová", "Florinel Hociung", "Petter Hohle", "Eva Huber", "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", "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", "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", "Petya Osenova", "Robert Östling", "Lilja Øvrelid", "Şaziye Betül Özateş", "Merve Özçelik", "Arzucan Özgür", "Balkız Öztürk Başaran", "Hyunji Hayley Park", "Niko Partanen", "Elena Pascual", "Marco Passarotti", "Agnieszka Patejuk", "Guilherme Paulino-Passos", "Angelika Peljak-Łapińska", "Siyao Peng", "Cenel-Augusto Perez", "Natalia Perkova", "Guy Perrier", "Slav Petrov", "Daria Petrova", "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", "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", "Steinþór Steingrímsson", "Antonio Stella", "Milan Straka", "Emmett Strickland", "Jana Strnadová", "Alane Suhr", "Yogi Lesmana Sulestio", "Umut Sulubacak", "Shingo Suzuki", "Zsolt Szántó", "Chihiro Taguchi", "Dima Taji", "Yuta Takahashi", "Fabio Tamburini", "Mary Ann C. Tan", "Takaaki Tanaka", "Dipta Tanaya", "Samson Tella", "Isabelle Tellier", "Marinella Testori", "Guillaume Thomas", "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", "Sowmya Vajjala", "Rob van der Goot", "Martine Vanhove", "Daniel van Niekerk", "Gertjan van Noord", "Viktor Varga", "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", "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
cs
2021
Forms and functions of of circumstantial modifications in Czech. Spatial and temporal modification
Formy a funkce okolnostních určení v češtině. Určení prostorová a časová
The necessity of a subdivision of the functors as bearers of the main semantic features in the Functional Generative Description as well as in the annotation scenario for the Prague Dependency Treebanks into more grained units called there subfunctors is presented here. The discussion is focused on the spatial and temporal functors as the most frequently occurring modification types. The book consists of the four parts. - Part I: Theoretical introduction to the description of adverbials analysis (repertory of functors, boundary between functors and subfunctors, principles of subfunctor determination, synonymy and ambiguity in form-function relation, determination of the class of secondary prepositions). - Part II: Corpus-based, detailed and comprehensive description of spatial and temporal modification - Part III: Two comparative studies. - Part IV: Overview of proposed functors and subfunctors
V monografii předkládáme celistvý popis forem a funkcí vybraných okolnostních určení založený na detailní analýze rozsáhlých korpusových dat. Usilujeme o komplexnost popisu co do šíře i co do hloubky: analyzujeme všechny v korpusech dostupné příklady a pro každý stanovujeme jeho funkci a formu. Výsledkem je podrobný seznam okolnostních významů a jejich jemnějších významových podtypů včetně seznamu formálních realizací, doložený reálnými příklady. Analýza se soustředí na popis určení prostorových a časových jakožto jádra okolnostních významů. Kniha se skládá ze čtyř částí. - Část I: Teoretický úvod do analýzy okolnostních určení (repertoár funktorů, hranice mezi funktory a subfunktory, principy determinace subfunktoru, synonymie a nejednoznačnost ve vztahu forma-funkce, stanovení třídy sekundárních předložek). - Část II: Podrobný a komplexní popis prostorových a časových určení - Část III: Dvě srovnávací studie. - Část IV: Přehled navržených funktorů a subfunktorů
[ "Marie Mikulová", "Jarmila Panevová" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
cs
2021
Valency Lexicon of Czech Verbs VALLEX
Valenční slovník českých sloves VALLEX
The Valency Dictionary of Czech Verbs (also VALLEX 4.0) provides information on the valency structure of Czech verbs in their individual meanings, characterized by glosses and examples; these data are supplemented by other syntactic and semantic characteristics. VALLEX 4.0 captures 4,659 Czech verbs, which correspond to 11,030 lexical units, "a given verb in a given meaning" (which corresponds to 6,829 lexical units if the mode counterparts are grouped into lexemes). VALLEX 4.0 enriches the information with the characteristics of verbs expressing reflexive and reciprocal meanings - these are both verbs that contain these features in their lexical meaning, and verbs expressing reciprocity or reflexivity using specific syntactic constructions. VALLEX 4.0 thus characterizes 2,909 verbs in 4,453 lexical units, which express or can express reciprocal relations (corresponding to 2,744 lexical units if the visual counterparts are grouped into lexemes), and 2,291 verbs in 3,287 lexical units with the possibility of expressing reflexivity (these are 2,039 lexical units for mode counterparts grouped into lexemes).
Předkládaný Valenční slovník českých sloves (též VALLEX 4.0) poskytuje informace o valenční struktuře českých sloves v jejich jednotlivých významech, které charakterizuje pomocí glos a příkladů; tyto údaje jsou doplněny o další syntaktické a sémantické charakteristiky. VALLEX 4.0 zachycuje 4 659 českých sloves, která odpovídají 11 030 lexikálním jednotkám, tedy vždy „danému slovesu v daném významu“ (což odpovídá 6 829 lexikálním jednotkám, sdruží-li se vidové protějšky do lexémů). VALLEX 4.0 poskytované informace obohacuje o charakteristiku sloves vyjadřujících reflexivní a reciproční významy – jde jednak o slovesa, která tyto rysy obsahují ve svém lexikálním významu, jednak o slovesa vyjadřující reciprocitu nebo reflexivitu pomocí specifických syntaktických konstrukcí. VALLEX 4.0 takto charakterizuje 2 909 sloves ve 4 453 lexikálních jednotkách, která vyjadřují či mohou vyjadřovat reciproční vztahy (což odpovídá 2 744 lexikálním jednotkám, sdruží-li se vidové protějšky do lexémů), a 2 291 sloves ve 3 287 lexikálních jednotkách s možností vyjadřovat reflexivitu (jde o 2 039 lexikálních jednotek pro vidové protějšky sdružené do lexémů).
[ "Markéta Lopatková", "Václava Kettnerová", "Anna Vernerová", "Eduard Bejček", "Zdeněk Žabokrtský" ]
https://www.semanticscholar.org/paper/975e45e9396bcdd6f76be4f89deccd900b57613f/
null
null
null
null
null
null
null
null
null
null
null
null
cs
2021
VALLEX dictionary and its practical use
Valenční slovník VALLEX a jeho praktické využití
In this text we present the VALLEX dictionary in both its components: data and rule. We assume that the methodology used here is useful for further lexicographic work. We also proposed a way to verify the suitability of the distinctions used in VALLEX on examples from corpora. With regard to the frequency of individual types of delimitation of meanings in texts, we also proposed certain simplifications on the scale of meanings on the example of the verb projet / projíždět as a process necessary for possible future modifications of passwords in the VALLEX dictionary.
V tomto textu na vybraných příkladech představujeme fungování slovníku VALLEX v obou jeho složkách: datové i pravidlové. Předpokládáme, že metodologie zde užitá je užitečná pro další lexikografickou práci. Navrhly jsme také způsob, jak na příkladech z korpusů ověřit vhodnost distinkcí užitých ve VALLEXu. S ohledem na frekvenci jednotlivých typů delimitace významů v textech jsme na příkladu slovesa projet / projíždět navrhly i jistá zjednodušení na škále významů jakožto proces potřebný pro případné budoucí úpravy hesel ve slovníku VALLEX.
[ "Markéta Lopatková", "Jarmila Panevová" ]
https://www.semanticscholar.org/paper/5ad34ef46cef8bdd6fc9d16a6541d4de466164a3/
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
From Raw Text to Enhanced Universal Dependencies: the Parsing Shared Task at IWPT 2021
Od prostého textu k Enhanced Universal Dependencies: soutěž v automatické syntaktické analýze na IWPT 2021
We describe the second IWPT task on end-to-end parsing from raw text to Enhanced Universal Dependencies. We provide details about the evaluation metrics and the datasets used for training and evaluation. We compare the approaches taken by participating teams and discuss the results of the shared task, also in comparison with the first edition of this task.
Popisujeme druhou soutěž IWPT v automatické analýze prostého textu do struktur Enhanced Universal Dependencies. Uvádíme podrobnosti o mírách použitých při vyhodnocení, jakož i o datových sadách použitých pro učení a vyhodnocení. Srovnáváme přístupy jednotlivých týmů, které se soutěže zúčastnily, a rozebíráme výsledky, mimo jiné i ve srovnání s výsledky prvního ročníku této soutěže.
[ "Gosse Bouma", "Djamé Seddah", "Daniel Zeman" ]
https://www.semanticscholar.org/paper/7226d14e6dea73dfad521256248ec2b19ae66ad8/
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
Artificial homonymy
Umělá homonymie
The paper presents a discussion about homonymy of Czech nouns with different or varying genders. The lemmas with this type of homonymy are treated in the new release of the dictionary MorfFlex as separated. We show that the separation of paradigms according to the gender is not only superfluous, but also clumsy, because it forces to make a choice when it is not necessary. That’s why we call this type of homonymy “artificial”.
Příspěvek přináší diskusi o homonymii českých podstatných jmen s různým nebo kolísavým rodem. Lemata s tímto typem homonymie jsou v novém vydání slovníku MorfFlex považována za různá. Ukazujeme, že rozdělení paradigmat podle rodu je nejen zbytečné, ale také nepraktické. Proto tomuto druhu hononymie říkáme „umělé“.
[ "Jaroslava Hlaváčová" ]
https://www.semanticscholar.org/paper/3ce3f931d0a0cc56e5061552b4c4c36e94a45caf/
null
null
null
null
null
null
null
null
null
null
null
null
cs
2021
Consistency of the MorfFlex morphological dictionary
Konzistence morfologického slovníku MorfFlex
Language corpora usually contain, in addition to their own texts, various types of annotations. The most common one is a morphological annotation, which consists in assigning a lemma and a morphological tag to each wordform. For morphological tagging, morphological dictionaries are traditionally used. Our paper presents a new version of the so-called “Prague” morphological dictionary MorfFlex used for tagging many Czech corpora (particularly Prague Dependency Treebanks, corpora published by the Institute of the Czech National Corpus in Prague or large Czech web corpora of the Aranea series). Three basic principles were used to update the dictionary: the Golden Rule of Morphology, the Principle of Paradigm Unity, and the Principle of Paradigm Uniqueness.
Jazykové korpusy obvykle obsahují kromě vlastních textů i různé typy anotací. Nejběžnější je morfologická anotace, která spočívá v přiřazení lemmatu a morfologického tagu každému slovnímu tvaru. Pro morfologické značkování se tradičně používají morfologické slovníky. Náš příspěvek představuje novou verzi tzv. „pražského“ morfologického slovníku MorfFlex používaného pro tagování mnoha českých korpusů (zejména Prague Dependency Treebanks, korpusy vydávané Ústavem českého národního korpusu v Praze nebo velké české webové korpusy řady Aranea ). K aktualizaci slovníku byly použity tři základní principy: Zlaté pravidlo morfologie, Princip jednoty paradigmatu a Princip jedinečnosti paradigmatu.
[ "Jaroslava Hlaváčová", "Marie Mikulová", "Barbora Štěpánková" ]
https://www.semanticscholar.org/paper/34209326060873e4019360847d2e453f9d642118/
null
null
null
null
null
null
null
null
null
null
null
null
cs
2021
Readability - Research and Practice
Srozumitelnost textu - Výzkum a praxe
Readability as (1) language competence expressed by CEFR levels, (2) literacy/reading comprehension level in OECD PIAAC studies, (3) a set of textual features. Benefits of clarity: time and cost savings, social cohesion, trust in public administration
Koncept textové srozumitelnosti: jako jazyková kompetence vyjádřená v úrovních CEFR, jako čtenářská gramotnost ve studiích OECD PIAAC, jako souhrn textových vlastností. Výhody, které přináší: úspora času, úspora peněz, společenská soudržnost, důvěra ve veřejné instituce.
[ "Silvie Cinková" ]
https://www.semanticscholar.org/paper/9267103b433ad6f1940d307cd6e638be1d8836c8/
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
Easy Language in Czechia
Snadné čtení (ETR) v Česku
The Handbook of Easy Languages in Europe describes the historical background, the principles and the practices of Easy Language in 21 European countries. The notion of Easy Language refers to modified forms of standard languages that aim to facilitate reading and comprehension.
Příručka snadných jazyků v Evropě popisuje historické pozadí, principy a postupy Easy Language ve 21 evropských zemích. Pojem Easy Language odkazuje na upravené formy standardních jazyků, jejichž cílem je usnadnit čtení a porozumění.
[ "Silvie Cinková", "Camille Latimier" ]
https://www.semanticscholar.org/paper/10c26c18248d9b6c7569cf4a674d4879b49a3b35/
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
EngVallex - English Valency Lexicon 2.0
EngVallex - valenční slovník angličtiny 2.0
EngVallex 2.0 as a slightly updated version of EngVallex. It is the English counterpart of the PDT-Vallex valency lexicon, using the same view of valency, valency frames and the description of a surface form of verbal arguments. EngVallex contains links also to PropBank (English predicate-argument lexicon). The EngVallex lexicon is fully linked to the English side of the PCEDT parallel treebank(s), which is in fact the PTB re-annotated using the Prague Dependency Treebank style of annotation. The EngVallex is available in an XML format in our repository, and also in a searchable form with examples from the PCEDT. EngVallex 2.0 is the same dataset as the EngVallex lexicon packaged with the PCEDT 3.0 corpus, but published separately under a more permissive licence, avoiding the need for LDC licence which is tied to PCEDT 3.0 as a whole.
EngVallex 2.0 je aktualizovaná verze slovníku EngVallex. Jedná se o anglický protějšek valenčního slovníku PDT-Vallex, který používá stejný pohled na valenci, valenční rámečky a popis povrchové formy slovních argumentů. EngVallex obsahuje také odkazy na PropBank (anglický predicate-argument lexicon). Slovník EngVallex je plně propojen s anglickou stranou PCEDT paralelního treebanku, což je ve reanotovaný PTB korpus ve stylu anotace Prague Dependency Treebank. EngVallex je k dispozici ve formátu XML a také ve formě k online vyhledávání s příklady z PCEDT. EngVallex 2.0 je stejný datový soubor jako EngVallex ve vydání PCEDT 3.0, ale vydává se samostatně na základě benevolentnější licence, čímž se předejde nutnosti pro uživatele získat LDC licenci, která je vázána na PCEDT 3.0 jako celek.
[ "Silvie Cinková", "Eva Fučíková", "Jana Šindlerová", "Jan Hajič" ]
https://www.semanticscholar.org/paper/0bc921c1895854ff97e297930c6ae7f3bfdeff03/
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
tidystopwords: Customisable Stop-Words in 110 Languages, v. 0.9.1
tidystopwords: přizpůsobitelné seznamy stopslov ve 110 jazycích, v. 0.9.1
Functions to generate stop-word lists in 110 languages, in a way consistent across all the languages supported. The generated lists are based on the morphological tagset from the Universal Dependencies.
Funkce na generování stop-slov v 110 jazycích, na základě morfologického značkování Universal Dependencies
[ "Silvie Cinková", "Maciej Eder" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
cs
2021
Baseline corpus for automatic simplification of Czech legal and administrative texts
Referenční datový soubor pro automatickou simplifikaci českých právních a úředních textů
Presentation of the Argumentation Mining task and an outline of the annotation scheme to train it, explained to law experts - potential text contributors and annotators. A broader context of the task, state of the art in NLP.
Představení úlohy Argumentation Mining a náčrt anotačního schématu k trénovacím datům z pohledu právních expertů - potenciálních přispěvatelů textů a anotátorů. Širší kontext úlohy, současné možnosti NLP.
[ "Silvie Cinková", "Jana Šamánková" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
Action nouns vs. nouns as bases for denominal verbs in Czech: A case study on directionality in derivation
Dějová substantiva vs. substantiva jako báze pro denominální slovesa v češtině: Případová studie o směrovosti v derivaci
Suffixless action nouns are mostly analysed as deverbal derivatives (e.g., výběr ‘choice’ < vybírat ‘to choose.IPFV’), but dictionaries ascribe the reverse direction to some noun–verb pairs (útok ‘attack’ > útočit ‘to attack.IPFV’) despite being both formally and semantically close to the former type. The question is addressed in the present study of whether any linguistic features can be identified in pairs of suffixless nouns and directly corresponding verbs that would speak in favour of one or the other direction. The analysis of 250 Czech suffixless nouns reveals a correlation between the number of directly related verbs derived by suffixes and the direction as recorded in the dictionaries: While deverbal nouns correspond mostly to a pair of verbs with different (aspect-changing) suffixes (cf. výběr ‘choice’ : vybrat/vybírat ‘to choose.PFV/IPFV’), nouns that are bases for verbs tend to share the root with a single (imperfective) verb (útok ‘attack’ : útočit ‘to attack.IPFV’). This correlation is elaborated into two different paradigms, one being based on verbal roots while the other on nominal roots, which might be applicable in hypothesizing the direction also with nouns that are not covered by the dictionaries.
Studie se zabývá českými substantivy s nulovou příponou. Poukazuje na to, že pokud se substantivum vztahuje ke dvěma slovesům s různou tematickou příponou, jedná se o substantivum od těchto sloves odvozené. Oproti tomu substantiva, pro která je v korpusových datech k dispozici sloveso s jedinou tematickou příponou a toto sloveso mění vid připojením předpony, slouží jako slova základová pro odvození (těchto denominálních) sloves.
[ "Magda Ševčíková" ]
https://www.semanticscholar.org/paper/5dc7b3f3c1bf6682e1ea74e9654bc3c06d0a1334/
null
null
null
null
null
null
null
null
null
null
null
null
cs
2021
Suffixless nouns and the category of grammatical aspect in related verbs
Bezpříponová substantiva a vyjadřování vidového protikladu u příbuzných sloves
The present study deals with suffixless nouns in Czech. Two data samples are analysed to demonstrate that suffixless nouns with an action meaning mostly correspond to a pair of verbs with different themes (conveying different grammatical aspects; e.g., skok ‘jump’ < skočit ‘to jump.PFV’ : skákat ‘to jump.IPFV’), whereas non-actional suffixless nouns tend to form a single corresponding verb which uses a prefix to change the aspect (noc ‘night’ > nocovat ‘to stay.IPFV overnight’ > přenocovat ‘to stay.PFV overnight’). This distinction is applied to a third data sample to determine direction of motivation in pairs of suffixless nouns and verbs. The difference between the deverbal and denominal direction is explained by the part-of-speech category of the root morpheme which is shared by the suffixless noun and the corresponding verb(s). The relations observed are modelled as paradigms recurring across the word-formation system of Czech.
Tato studie se zabývá českými podstatnými jmény bez slovotvorné přípony. Na dvou vzorcích dat je analyzováno, že bezpříponová substantiva s dějovým významem většinou odpovídají dvojici sloves s různými kmenotvornými příponami (vyjadřujícími různé vidové hodnoty; např. skok < skočit : skákat), zatímco bezpříponový substantiva s jiným než dějovým významem zpravidla korespondují s jediným slovesem, které nemění vid změnou kmentovorné přípony, ale připojením předpony (noc > nocovat > přenocovat). Způsob tvoření vidového protějšku je zkoumán ve třetím vzorku dat s cílem určit v příslušných dvojicích dvojic bezpříponového substantiva a slovesa směr motivace. Rozdíl mezi deverbálním a denominálním směrem je spojován se slovnědruhovou kategorií kořenového morfému, který je společný pro bezpříponové substantivum a příslušné sloveso (slovesa). Pozorované vztahy jsou modelovány jako paradigmata opakující se ve slovotvorném systému češtiny.
[ "Magda Ševčíková" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
cs
2021
The DeriNet lexical database as a formal model of word formation in Czech
Lexikální databáze DeriNet jako formální model tvoření slov v češtině
In this talk, we will summarize the main features of the word formation network DeriNet developed for Czech at the Institute of Formal and Applied Linguistics, MFF UK. We will also discuss two topics that have been recently addressed in connection with DeriNet. The first is related to the applicability of the so-called paradigmatic approach to Czech word formation, in contrast with the tree approach taken in DeriNet. The second topic is the segmentation of Czech lemmas into morphemes.
V přednášce shrneme hlavní vlastnosti slovotvorné sítě DeriNet vyvíjené pro češtinu v Ústavu formální a aplikované lingvistiky MFF UK. Nastíníme rozmanitost podobně zaměřených databází vznikajících pro jiné jazyky, jakož i neskromné vyhlídky na jejich syntézu ve stopách Universal Dependencies. Dále se budeme věnovat dvěma tématům, kterými se v souvislosti s DeriNetem zabýváme v poslední době. První souvisí s aplikovatelností tzv. paradigmatického přístupu k odvozování v češtině, kontrastovaného se stromovým přístupem zvoleným v DeriNetu. Druhým tématem je segmentace českých lemmat na morfémy; jakkoli je morfém jedním z “nejklasičtějších” lingvistických pojmů, jeho rigorózní datové uchopení se zdá být překvapivě nesnadnou výzvou.
[ "Magda Ševčíková", "Zdeněk Žabokrtský" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
Agent noun formation in Czech: An empirical study on suffix rivalry
Tvoření agentních jmen v češtině: empirická studie kompetice přípon
The work deals with the formation of agent nouns in Czech, especially the rivalry among the eight most frequent agent suffixes. We use machine learning methods to calculate the influence of several different formal-linguistic properties on the resulting suffix selected when an agent noun is coined.
Práce se zabývá tvorbou agentních jmen v češtině, zejména kompeticí mezi osmi nejfrekventovanějšími agentními příponami. Na základě metod strojového učení je v práci vyčíslen vliv různých formálně-lingvistických vlastností na výslednou volbu přípony při tvorbě agentního substantiva.
[ "Magda Ševčíková", "Lukáš Kyjánek", "Barbora Vidová Hladká" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
ParlaMint: Comparable Corpora of European Parliamentary Data
ParlaMint: Porovnatelné korpusy evropských parlamentních dat
This paper outlines the ParlaMint project from the perspective of its goals, tasks, participants, results and applications potential. The project produced language corpora from the sessions of the national parliaments of 17 countries, almost half a billion words in total. The corpora are split into COVID-related subcorpora (from November 2019) and reference corpora (to October 2019). The corpora are uniformly encoded according to the ParlaMint schema with the same Universal Dependencies linguistic annotations. Samples of the corpora and conversion scripts are available from the project’s GitHub repository. The complete corpora are openly available via the CLARIN.SI repository for download, and through the NoSketch Engine and KonText concordancers as well as through the Parlameter interface for exploration and analysis.
Tento článek popisuje projekt ParlaMint z hlediska jeho cílů, úkolů, účastníků, výsledků a aplikačního potenciálu. Projekt vytvořil jazykové korpusy ze zasedání národních parlamentů 17 zemí, celkem téměř půl miliardy slov. Korpusy jsou rozděleny na subkorpusy související s COVID (od listopadu 2019) a referenční korpusy (do října 2019). Korpusy jsou jednotně kódovány podle schématu ParlaMint se stejnými lingvistickými anotacemi podle Universal Dependencies. Ukázky korpusů a konverzních skriptů jsou dostupné z GitHub úložiště projektu. Kompletní korpusy je volně dostupné ke stažení přes repozitář CLARIN.SI a přes concordancery NoSketch Engine a KonText i přes rozhraní Parlameter pro procházení a analýzu.
[ "Tomaž Erjavec", "Maciej Ogrodniczuk", "Petya Osenova", "Andrej Pančur", "Nikola Ljubešić", "Tommaso Agnoloni", "Starkaður Barkarson", "María Calzada Pérez", "Çağrı Çöltekin", "Matthew Coole", "Roberts Darģis", "Luciana de Macedo", "Jesse de Does", "Katrien Depuydt", "Sascha Diwersy", "Dorte Haltrup Hansen", "Matyáš Kopp", "Tomas Krilavičius", "Giancarlo Luxardo", "Maarten Marx", "Vaidas Morkevičius", "Costanza Navarretta", "Paul Rayson", "Orsolya Ring", "Michał Rudolf", "Kiril Simov", "Steinþór Steingrímsson", "István Üveges", "Ruben van Heusden", "Giulia Venturi" ]
https://www.semanticscholar.org/paper/382c276d6084fc2a8a9963e1806c5b505f20a07a/
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
Multilingual comparable corpora of parliamentary debates ParlaMint 2.1
Vícejazyčný korpus parlamentních debat ParlaMint 2.1
ParlaMint is a multilingual set of comparable corpora containing parliamentary debates. It contains parliament debates from these countries: Belgium, Bulgaria, Czech Republic, Denmark, Croatia, France, Iceland, Italy, Lithuania, Latvia, Hungary, Netherlands, Poland, Slovenia, Spain, Turkey, UK
ParlaMint je vícejazyčný soubor srovnatelných korpusů parlamntních debat. Obsahuje parlamentní data těchto zemí: Belgie, Bulharsko, Česko, Dánsko, Francie, Chorvatsko, Island, Itálie, Litva, Lotyšsko, Maďarsko, Nizozemí, Polsko, Slovinsko, Španělsko, Turecko, UK.
[ "Tomaž Erjavec", "Maciej Ogrodniczuk", "Petya Osenova", "Nikola Ljubešić", "Kiril Simov", "Vladislava Grigorova", "Michał Rudolf", "Andrej Pančur", "Matyáš Kopp", "Starkaður Barkarson", "Steinþór Steingrímsson", "Henk van der Pol", "Griet Depoorter", "Jesse de Does", "Bart Jongejan", "Dorte Haltrup Hansen", "Costanza Navarretta", "María Calzada Pérez", "Luciana de Macedo", "Ruben van Heusden", "Maarten Marx", "Çağrı Çöltekin", "Matthew Coole", "Tommaso Agnoloni", "Francesca Frontini", "Simonetta Montemagni", "Valeria Quochi", "Giulia Venturi", "Manuela Ruisi", "Carlo Marchetti", "Roberto Battistoni", "Miklós Sebők", "Orsolya Ring", "Roberts Darģis", "Andrius Utka", "Mindaugas Petkevičius", "Tomas Krilavičius", "Vaidas Morkevičius", "Sascha Diwersy", "Giancarlo Luxardo" ]
https://www.semanticscholar.org/paper/7f8287dfa5e19c9bdce0895eaf529b46d74ea391/
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
Multilingual comparable corpora of parliamentary debates ParlaMint 2.0
Vícejazyčný korpus parlamentních debat ParlaMint 2.0
ParlaMint is a multilingual set of comparable corpora containing parliamentary debates. It contains parliament debates from these countries: Belgium, Bulgaria, Czech Republic, Denmark, Croatia, Iceland, Italy, Lithuania, Latvia, Hungary, Netherlands, Poland, Slovenia, Spain, Turkey, UK
ParlaMint je vícejazyčný soubor srovnatelných korpusů parlamntních debat. Obsahuje parlamentní data těchto zemí: Belgie, Bulharsko, Česko, Dánsko, Chorvatsko, Island, Itálie, Litva, Lotyšsko, Maďarsko, Nizozemí, Polsko, Slovinsko, Španělsko, Turecko, UK.
[ "Tomaž Erjavec", "Maciej Ogrodniczuk", "Petya Osenova", "Nikola Ljubešić", "Kiril Simov", "Vladislava Grigorova", "Michał Rudolf", "Andrej Pančur", "Matyáš Kopp", "Starkaður Barkarson", "Steinþór Steingrímsson", "Henk van der Pol", "Griet Depoorter", "Jesse de Does", "Bart Jongejan", "Dorte Haltrup Hansen", "Costanza Navarretta", "María Calzada Pérez", "Luciana de Macedo", "Ruben van Heusden", "Maarten Marx", "Çağrı Çöltekin", "Matthew Coole", "Tommaso Agnoloni", "Francesca Frontini", "Simonetta Montemagni", "Valeria Quochi", "Giulia Venturi", "Manuela Ruisi", "Carlo Marchetti", "Roberto Battistoni", "Miklós Sebők", "Orsolya Ring", "Roberts Darģis", "Andrius Utka", "Mindaugas Petkevičius", "Monika Briedienė", "Tomas Krilavičius", "Vaidas Morkevičius" ]
https://www.semanticscholar.org/paper/7f8287dfa5e19c9bdce0895eaf529b46d74ea391/
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
Linguistically annotated multilingual comparable corpora of parliamentary debates ParlaMint.ana 2.0
Lingvisticky anotovaný vícejazyčný korpus parlamentních debat ParlaMint.ana 2.0
ParlaMint.ana is a multilingual set of comparable linguistically annotated corpora containing parliamentary debates. It contains parliament debates from these countries: Belgium, Bulgaria, Czech Republic, Denmark, Croatia, Iceland, Italy, Lithuania, Latvia, Hungary, Netherlands, Poland, Slovenia, Spain, Turkey, UK
ParlaMint.ana je vícejazyčný soubor srovnatelných lingvisticky anotovaných korpusů parlamntních debat. Obsahuje parlamentní data těchto zemí: Belgie, Bulharsko, Česko, Dánsko, Chorvatsko, Island, Itálie, Litva, Lotyšsko, Maďarsko, Nizozemí, Polsko, Slovinsko, Španělsko, Turecko, UK.
[ "Tomaž Erjavec", "Maciej Ogrodniczuk", "Petya Osenova", "Nikola Ljubešić", "Kiril Simov", "Vladislava Grigorova", "Michał Rudolf", "Andrej Pančur", "Matyáš Kopp", "Starkaður Barkarson", "Steinþór Steingrímsson", "Henk van der Pol", "Griet Depoorter", "Jesse de Does", "Bart Jongejan", "Dorte Haltrup Hansen", "Costanza Navarretta", "María Calzada Pérez", "Luciana de Macedo", "Ruben van Heusden", "Maarten Marx", "Çağrı Çöltekin", "Matthew Coole", "Tommaso Agnoloni", "Francesca Frontini", "Simonetta Montemagni", "Valeria Quochi", "Giulia Venturi", "Manuela Ruisi", "Carlo Marchetti", "Roberto Battistoni", "Miklós Sebők", "Orsolya Ring", "Roberts Darģis", "Andrius Utka", "Mindaugas Petkevičius", "Monika Briedienė", "Tomas Krilavičius", "Vaidas Morkevičius", "Roberto Bartolini", "Andrea Cimino" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
Linguistically annotated multilingual comparable corpora of parliamentary debates ParlaMint.ana 2.1
Lingvisticky anotovaný vícejazyčný korpus parlamentních debat ParlaMint.ana 2.1
ParlaMint.ana is a multilingual set of comparable linguistically annotated corpora containing parliamentary debates. It contains parliament debates from these countries: Belgium, Bulgaria, Czech Republic, Denmark, France, Croatia, Iceland, Italy, Lithuania, Latvia, Hungary, Netherlands, Poland, Slovenia, Spain, Turkey, UK
ParlaMint.ana je vícejazyčný soubor srovnatelných lingvisticky anotovaných korpusů parlamntních debat. Obsahuje parlamentní data těchto zemí: Belgie, Bulharsko, Česko, Dánsko, Francie, Chorvatsko, Island, Itálie, Litva, Lotyšsko, Maďarsko, Nizozemí, Polsko, Slovinsko, Španělsko, Turecko, UK.
[ "Tomaž Erjavec", "Maciej Ogrodniczuk", "Petya Osenova", "Nikola Ljubešić", "Kiril Simov", "Vladislava Grigorova", "Michał Rudolf", "Andrej Pančur", "Matyáš Kopp", "Starkaður Barkarson", "Steinþór Steingrímsson", "Henk van der Pol", "Griet Depoorter", "Jesse de Does", "Bart Jongejan", "Dorte Haltrup Hansen", "Costanza Navarretta", "María Calzada Pérez", "Luciana de Macedo", "Ruben van Heusden", "Maarten Marx", "Çağrı Çöltekin", "Matthew Coole", "Tommaso Agnoloni", "Francesca Frontini", "Simonetta Montemagni", "Valeria Quochi", "Giulia Venturi", "Manuela Ruisi", "Carlo Marchetti", "Roberto Battistoni", "Miklós Sebők", "Orsolya Ring", "Roberts Darģis", "Andrius Utka", "Mindaugas Petkevičius", "Monika Briedienė", "Tomas Krilavičius", "Vaidas Morkevičius", "Roberto Bartolini", "Andrea Cimino", "Sascha Diwersy", "Giancarlo Luxardo" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
More Data and New Tools. Advances in Parsing the Index Thomisticus Treebank
Více dat a nové nástroje. Pokroky v parsování Index Thomisticus Treebank
This paper investigates the recent advances in parsing the Index Thomisticus Treebank, which encompasses Medieval Latin texts by Thomas Aquinas. The research focuses on two types of variables. On the one hand, it examines the impact that a larger dataset has on the results of parsing; on the other hand, performances of new parsers are analysed with respect to less recent tools. Term of comparison to determine the effective parsing advances are the results in parsing the Index Thomisticus Treebank described in a previous work. First, the best performing parser among those concerned in that study is tested on a larger dataset than the one originally used. Then, some parser combinations that were developed in the same study are evaluated as well, assessing that more training data result in more accurate performances. Finally, to examine the impact that newly available tools have on parsing results, we train, test, and evaluate two neural parsers chosen among those best performing in the CoNLL 2018 Shared Task. Our experiments reach the highest accuracy rates achieved so far in automatic syntactic parsing of the Index Thomisticus Treebank and of Latin overall.
Tento článek zkoumá nedávné pokroky v analýze Index Thomisticus Treebank, který zahrnuje středověké latinské texty Tomáše Akvinského. Výzkum se zaměřuje na dva typy proměnných. Na jedné straně zkoumá, jaký vliv má větší soubor dat na výsledky parsování, na druhé straně jsou analyzovány výkony nových parserů s ohledem na méně aktuální nástroje. Termínem srovnání pro určení efektivního pokroku v parsování jsou výsledky při parsování Index Thomisticus Treebank popsané v předchozí práci. Nejprve je nejvýkonnější parser z těch, kterých se týkala tato studie, testován na větším souboru dat, než byl ten původně použitý. Poté jsou vyhodnoceny i některé kombinace parserů, které byly vyvinuty v téže studii, přičemž je posouzeno, že více trénovacích dat vede k přesnějším výkonům. Nakonec, abychom prozkoumali, jaký vliv mají nově dostupné nástroje na výsledky parsování, trénujeme, testujeme a vyhodnocujeme dva neuronové parsery vybrané mezi těmi, které dosáhly nejlepších výsledků ve sdílené úloze CoNLL 2018. Naše experimenty dosahují dosud nejvyšší dosažené míry přesnosti v automatickém syntaktickém rozboru Index Thomisticus Treebank a latiny celkově.
[ "Federica Gamba", "Marco Passarotti", "Paolo Ruffolo" ]
https://www.semanticscholar.org/paper/8577d80f2df6cba4cdae00630661d5d31028c7f3/
null
null
null
null
null
null
null
null
null
null
null
null
cs
2021
computer science seminar "THEaiTRE project"
informatický seminář „Projekt THEaiTRE“
We've been hearing more and more about artificial intelligence lately. It plays chess, translates texts, drives cars... But can artificial intelligence create art, such as writing a play? What can we expect from it in the future? What areas of human life will it interfere with? And what exactly is this artificial intelligence?
O umělé inteligenci slyšíme v poslední době víc a víc. Hraje šachy, překládá texty, řídí auta... Ale zvládne umělá inteligence tvořit umění, například napsat divadelní hru? Co od ní v budoucnosti můžeme očekávat? Do jakých oblastí lidského života zasáhne? A co to vlastně je, ta umělá inteligence?
[ "Rudolf Rosa" ]
https://www.semanticscholar.org/paper/bc30e61766ed68f9bbd1f52630ea3246a3ecfce3/
null
null
null
null
null
null
null
null
null
null
null
null
cs
2021
THEaiTRE project: artificial intelligence and art
Projekt THEaiTRE: umělá inteligence a umění
We've been hearing more and more about artificial intelligence lately. It plays chess, translates texts, drives cars... But can artificial intelligence create art, such as writing a play? What can we expect from it in the future? What areas of human life will it interfere with? And what exactly is this artificial intelligence?
O umělé inteligenci slyšíme v poslední době víc a víc. Hraje šachy, překládá texty, řídí auta... Ale zvládne umělá inteligence tvořit umění, například napsat divadelní hru? Co od ní v budoucnosti můžeme očekávat? Do jakých oblastí lidského života zasáhne? A co to vlastně je, ta umělá inteligence?
[ "Rudolf Rosa" ]
https://www.semanticscholar.org/paper/654aaf71b757fe87c4d2f0c04d02a0744c737216/
null
null
null
null
null
null
null
null
null
null
null
null
en
2021
Soft aspects of the THEaiTRE project
Měkké aspekty projektu THEaiTRE
Soft aspects of THEaiTRE (humanities, PR, discussions, spectators...)
Měkké aspekty projektu THEaiTRE (humanitní vědy, PR, diskuse, diváci...)
[ "Rudolf Rosa" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
cs
2021
Technical report about the development of the THEaiTRE project in 2020
Technická zpráva o vývoji projektu THEaiTRE v roce 2020
This project, combining computational linguistics and theatre, aims to directly reflect the onset and explore the role of artificial intelligence in society and art and the possibility of its cooperation with humans. We propose to combine theatrological research with computational linguistic research: the application of results from these fields will lead to the creation and training of a system for automatic generation of theatre play scripts. This system will be interactively used by our team to create a play that will be put on by a theatre on the occasion of the 100th anniversary of the R.U.R drama premiere. The resulting work of art and its acceptance by viewers will be examined, and the knowledge gained will be used to improve the system and to put on another play.
Tento projekt kombinující počítačovou lingvistiku a divadelnictví má za cíl přímým způsobem reflektovat nástup a zkoumat roli umělé inteligence ve společnosti a v umění a možnost její spolupráce s člověkem. Navrhujeme spojit teatrologický výzkum s komputačně lingvistickým výzkumem, kde aplikace poznatků a výsledků z těchto oborů povede k sestavení a natrénování systému pro automatické generování scénářů divadelních her. Tento systém následně bude naším týmem interaktivně využit pro vytvoření hry, která bude realizována hereckým souborem ke 100. výročí premiéry dramatu R.U.R. Vzniklé dílo a jeho přijetí diváky budeme zpětně zkoumat v rámci obou vědních oborů a získané poznatky využijeme pro další zdokonalení systému, který zpřístupníme na webu, a pro realizaci dalších her.
[ "Rudolf Rosa" ]
null
null
null
null
null
null
null
null
null
null
null
null
null