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
2018
Universal Dependencies and Non-Native Czech
Anotace češtiny nerodilých mluvčích pomocí formalismu UD
CzeSL is a learner corpus of texts produced by non-native speakers of Czech. Such corpora area great source of information about specific features of learners’ language, helping language teachers and researchers in the area of second language acquisition. In our project, we have focused on syntactic annotation of the non-native text within the framework of Universal Dependencies. As far as we know, this is a first project annotating a richly inflectional non-native language. Our ideal goal has been to annotate according to the non-native grammar in the mind of the author, not according to the standard grammar. However, this brings many challenges. First, we do not have enough data to get reliable insights into the grammar of each author. Second, many phenomena are far more complicated than they are in native languages. We believe that the most important result of this project is not the actual annotation, but the guidelines and principles that can be used as a basis for other non-native languages.
CzeSL je žákovský korpus textů nerodilých mluvčích češtiny, který je cenným zdrojem unikátních znalostí o jazyce nerodilých mluvčích. Pomáhá pedagogům a výzkumníkům v oblasti osvojování druhého jazyka. V našem projektu se zaměřujeme na syntaktickou anotaci textů z CzeSL v rámci formalismu Universal Dependencies. Pokud je nám známo, jedná se o první pokus takové anotace pro jazyk s bohatou flexí. Naší ambicí je anotovat dle gramatiky, kterou používá nerodilý mluvčí, namísto standardní gramatiky. Relativné malé množství dat přináší celou řadu otázek.
[ "Jirka Hana", "Barbora Hladká" ]
https://www.semanticscholar.org/paper/243c6d33ea433f69047602b26b0688676bf31323/
null
null
null
null
null
null
null
null
null
null
null
null
cs
2018
Olympiad in Czech language: 44th year
Olympiáda v českém jazyce: 44. ročník
The article describes the course of the 44th year of the Olympiad in the Czech Language, presenting its general settings as well as the tasks, their solutions, the approaches of the participants and the names of the winners.
Článek popisuje průběh 44. ročníku Olympiády v českém jazyce. Představuje celkový průběh soutěže, soutěžní úkoly a jejich řešení, komentuje řešení účastníků soutěže a přináší jména vítězů celostátního kola.
[ "Kateřina Rysová" ]
https://www.semanticscholar.org/paper/91d000166eafb80b9280e5eb038eb02d9c49f5e0/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
Discourse Connectives and Reference
Diskurzní konektory a reference
We examine discourse connectives from the perspective of reference (i.e. a presence of an anaphoric/cataphoric element). We introduce a division of connectives into: i) connectives without an inherent (internal) reference (e.g. "and", "but", "or", "if", "however", "then"), and ii) connectives with an inherent (internal) reference that is either optional (e.g. "as a result" vs. "as a result of this"), or obligatory – cf. already grammaticalized connectives (i.e. primary connectives like "thereafter", "therefore" or "thereby") vs. not yet grammaticalized connectives (i.e. secondary connectives like "because of this", "for this reason").
Diskurzní konektory představujeme z pohledu reference (tj. přítomnosti anaforického nebo kataforického prvku). Diskurzní konektory dělíme na: 1) konektory bez interní reference (např. "a", "ale", "nebo", "jestli", "však", "pak") a 2) konektory s interní referencí, která může být fakultativní (např. "výsledkem je" vs. "výsledkem toho je") nebo obligatorní – srov. již gramatikalizované konektory (tj. primární konektory typu "potom", "proto", "tímto") i dosud ne zcela gramatikalizované konektory (tj. sekundární konektory typu "kvůli tomu", "z tohoto důvodu").
[ "Kateřina Rysová", "Magdaléna Rysová" ]
https://www.semanticscholar.org/paper/018d47c42573db86cc937c3a2e7f3562a13c25b9/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
The Correlation between Discourse-Anaphoric Devices and an Overall Communicative Competence in Learners‘ Essays
Vztah mezi anaforickými diskurzními konektory a celkovou komunikační kompetencí ve studentských esejích
We introduce a contrastive study of text coherence in Czech and German. Specifically, we focus on the discourse-anaphoric devices (called anaphoric connectives) contributing to text coherence and we analyze their role in the overall communicative competence of both native and non-native speakers of Czech and German.
Článek se zabývá analýzou koherence textu v češtině a v němčině. Zaměřuje se na tzv. anaforické diskurzní konektory, které se významně podílejí na stavbě textu. Příspěvek analyzuje roli těchto konektorů v celkové komunikační kompetenci rodilých i nerodilých mluvčích češtiny a němčiny.
[ "Kateřina Rysová", "Magdaléna Rysová" ]
https://www.semanticscholar.org/paper/dafbd53ef06735c2f9e2c0e33fce466d6b2a3407/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
Primary and secondary discourse connectives: definitions and lexicons
Primární a sekundární konektory: definice a slovníky
Starting from the perspective that discourse structure arises from the presence of coherence relations, we provide a map of linguistic discourse structuring devices (DRDs), and then focus on those found in written text: connectives. To subdivide this class further, we follow the recent idea of structuring the set of connectives by differentiating between primary and secondary connectives, on the one hand, and free connecting phrases, on the other. Considering examples from Czech, English, French and German, we develop definitions of these groups, with attention to certain cross-linguistic differences. For primary and secondary connectives, we propose that their behavior can be described to a large extent by declarative lexicons, and we demonstrate a concrete proposal which has been applied to five languages, with others currently being added in ongoing work. The lexical representations can be useful both for humans (theoretical investigations, transfer to other languages) and for machines (automatic discourse parsing and generation).
Koherentní text (diskurz) se skládá ze vzájemně propojených vztahů, které mohou být vyjádřeny explicitními jazykovými prostředky. V našem článku se zabýváme jedním typem těchto prostředků, diskurzními konektory. Přebíráme přitom dělení konektivních prostředků na primární konektory, sekundární konektory a volné konektivní fráze. Na základě dokladů z češtiny, angličtiny, francouzštiny a němčiny rozvíjíme definice těchto skupin, především s ohledem na mezijazykové rozdíly. U primárních a sekundárních konektorů navrhujeme způsob jejich zachycení ve slovnících. Představujeme konkrétní návrh vícejazyčného slovníku konektorů, do kterého je v současné době zapojeno pět různých jazyků. Další jazyky budou postupně přidávány.
[ "Laurence Danlos", "Kateřina Rysová", "Magdaléna Rysová", "Manfred Stede" ]
https://www.semanticscholar.org/paper/14dbc964728ba6e71fe82b63ad7909bd9c217c82/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
Translating Short Segments with NMT: A Case Study in English-to-Hindi
Překlad krátkých segmentů pomocí NMT: Případová studie pro překlad z angličtiny do hindštiny
This paper presents a case study in translating short image captions of the Visual Genome dataset from English into Hindi using out-of-domain data sets of varying size. We experiment with three NMT models: the shallow and deep sequence-to-sequence and the Transformer model as implemented in Marian toolkit. Phrase-based Moses serves as the baseline. The results indicate that the Transformer model outperforms others in the large data setting in a number of automatic metrics and manual evaluation, and it also produces the fewest truncated sentences. Transformer training is however very sensitive to the hyperparameters, so it requires more experimenting. The deep sequence-to-sequence model produced more flawless outputs in the small data setting and it was generally more stable, at the cost of more training iterations.
This paper presents a case study in translating short image captions of the Visual Genome dataset from English into Hindi using out-of-domain data sets of varying size. We experiment with three NMT models: the shallow and deep sequence-to-sequence and the Transformer model as implemented in Marian toolkit. Phrase-based Moses serves as the baseline. The results indicate that the Transformer model outperforms others in the large data setting in a number of automatic metrics and manual evaluation, and it also produces the fewest truncated sentences. Transformer training is however very sensitive to the hyperparameters, so it requires more experimenting.
[ "Shantipriya Parida", "Ondřej Bojar" ]
https://www.semanticscholar.org/paper/2caaba23f9695a7ab5872cb3b93dae7ffae044f0/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
Morphological and Language-Agnostic Word Segmentation for NMT
Morfologická a jazykově nezávislá segmentace slov pro NMT
The state of the art of handling rich morphology in neural machine translation (NMT) is to break word forms into subword units, so that the overall vocabulary size of these units fits the practical limits given by the NMT model and GPU memory capacity. In this paper, we compare two common but linguistically uninformed methods of subword construction (BPE and STE, the method implemented in Tensor2Tensor toolkit) and two linguistically-motivated methods: Morfessor and one novel method, based on a derivational dictionary. Our experiments with German-to-Czech translation, both morphologically rich, document that so far, the non-motivated methods perform better. Furthermore, we identify a critical difference between BPE and STE and show a simple pre-processing step for BPE that considerably increases translation quality as evaluated by automatic measures.
V současných neuronových systémech pro strojový překlad textů přirozeného jazyka (NMT) se morfologicky příbuzná slova zpracovávají jejich rozdělením na podslovní jednotky takovým způsobem, aby se slovník jednotek vešel do limitů daných zvoleným NMT modelem a do paměti grafické karty. V tomto článku srovnáváme dva nejobvyklejší, nelingvistické, způsoby vytváření podslovních jednotek (BPE a STE, metody implementované v nástroji Tensor2Tensor) se dvěma lingvisticky motivovanými způsoby: Nástrojem Morfessor a námi vyvinutou metodou založenou na derivačních vztazích. Naše experimenty s překladem z němčiny do češtiny, morfologicky bohatých jazyků, ukazují, že prozatím mají lepší výsledky nelingvistické metody. K tomu identifikujeme důležitý rozdíl mezi BPE a STE a ukazujeme, že jednoduché předzpracování textu před BPE výrazně zvyšuje kvalitu překladu vyhodnocovanou automatickými metrikami.
[ "Dominik Macháček", "Jonáš Vidra", "Ondřej Bojar" ]
https://www.semanticscholar.org/paper/19885d5b288e6043eb989d296cc12bc8dbead8e3/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
NMT System Comparison
Porovnání systémů pro neuronový strojový překlad
This work is a comparison of a quality and speed of NMT systems Tensor2Tensor, Marian, Nematus, Neural Monkey and OpenNMT on two comparable translation tasks.
Tato práce porovnává kvalitu a rychlost systémů pro neuronový strojový překlad (Tensor2Tensor, Marian, Nematus, Neural Monkey, OpenNMT) na dvou srovnatelných překladových úlohách.
[ "Dominik Macháček", "Ondřej Bojar", "Martin Popel", "Tom Kocmi" ]
https://www.semanticscholar.org/paper/9b0580d9e6dc8f8cefad05f0e6198bcca0c2e608/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
Extracting Syntactic Trees from NMT Encoder Self-Attentions
Extrakce syntaktických struktur z self-attentions NMT enkodéru
Extracting Syntactic Trees from NMT Encoder Self-Attentions. Extracting Syntactic Trees from NMT Encoder Self-Attentions.
Extrakce syntaktických struktur z self-attentions NMT enkodéru. Extrakce syntaktických struktur z self-attentions NMT enkodéru.
[ "David Mareček", "Rudolf Rosa" ]
https://www.semanticscholar.org/paper/49400b3a3ea01772e321e3e010b7b891c3d6cb88/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
Extracting Syntactic Trees from Transformer Encoder Self-Attentions
Extrakce syntaktických struktur z self-attentions enkodéru Transformeru
Extracting Syntactic Trees from Transformer Encoder Self-Attentions. Extracting Syntactic Trees from Transformer Encoder Self-Attentions.
Extrakce syntaktických struktur z self-attentions enkodéru Transformeru. Extrakce syntaktických struktur z self-attentions enkodéru Transformeru
[ "David Mareček", "Rudolf Rosa" ]
https://www.semanticscholar.org/paper/49400b3a3ea01772e321e3e010b7b891c3d6cb88/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
Grounded Sequence-to-Sequence Transduction
Multimodální převádění sekvencí na sekvence
This presentation summarizes output of the Grounded Sequence-to-Sequence Transduction team at the Fifth Fred Jelinek Memorial Summer Workshop. We combine written, spoken, and seen objects and actions in how-to videos: if we see a person slicing round, red objects and putting them on a brown surface, it is more likely that she or he is explaining how to make a sandwich than how to change a car tire. And we might learn that the red objects are called “tomatoes”. Our team develop methods that exploit multimodality to process and analyze videos to accomplish three main tasks: speech captioning, video-to-text summarization and translation into a different language. These tasks are diverse but not unrelated. Therefore, we model them using a multi-task sequence-to-sequence learning framework where these (and other, auxiliary) tasks can benefit from shared representations.
Tato prezentace shrnuje závěry týmu Grounded Sequence-to-Sequence Transduction na pátém Fred Jelinek Memorial Summer Workshop. V prezentovaných modelech kombinujeme psaný text, mluvené slovo a objekty a akce v pozorované . Když vidíme, že člověk krájí červené předměty a umístí je na hnědý povrch, je pravděpodobné, že spíše vysvětluje, jak vyrobit sendvič, než jak změnit pneumatiku. A- mohli bychom se dozvědět, že červené předměty se nazývají "rajčata". Náš tým pracoval na vývoji metod, které využívají multimodální informace pro zpracování a analýzu videoklipů, ve třech hlavních lohách: rozpoznávání řeči, sumarizace videa a textu a překlad do jiného jazyka.
[ "Loïc Barrault", "Ozan Caglayan", "Amanda Duarte", "Desmond Elliott", "Spandana Gella", "Nils Holzenberger", "Chiraag Lala", "Sun Jae Lee", "Jindřich Libovický", "Pranava Madhyastha", "Florian Metze", "Karl Mulligan", "Alissa Ostapenko", "Shruti Palaskar", "Ramon Sanabria", "Lucia Specia", "Josiah Wang" ]
https://www.semanticscholar.org/paper/14aadd24040edaa9ce9978b53b00aeede015f859/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
Using Adversarial Examples in Natural Language Processing
Použití adversálních příkladů ve zpracování přirozeného jazyka
Machine learning models have been providing promising results in many fields including natural language processing. These models are, nevertheless, prone to adversarial examples. These are artificially constructed examples which evince two main features: they resemble the real training data but they deceive already trained model. This paper investigates the effect of using adversarial examples during the training of recurrent neural networks whose text input is in the form of a sequence of word/character embeddings. The effects are studied on a compilation of eight NLP datasets whose interface was unified for quick experimenting. Based on the experiments and the dataset characteristics, we conclude that using the adversarial examples for NLP tasks that are modeled by recurrent neural networks provides a regularization effect and enables the training of models with greater number of parameters without overfitting. In addition, we discuss which combinations of datasets and model settings might benefit from the adversarial training the most.
Modely strojového učení přinášejí slibné výsledky v mnoha oborech včetně zpracování přirozeného jazyka. Tyto modely jsou nicméně náchylné k adversálním příkladům. Jedná se o uměle vytvořené příklady s dvěma důležitými vlastnostmi: podobají se skutečným tréninkovým příkladům, ale matou již natrénovaný model. Tento článek zkoumá účinek používání adversálních příkladů při tréninku rekurentních neuronových sítí, jejichž vstup je ve formě slovních či znakových embeddingů. Účinky jsou studovány na kompilaci osmi datasetů. Na základě experimentů a charakteristik datasetů dospíváme k závěru, že použití adversálních příkladů pro úkoly zpracování přirozeného jazyka, které jsou modelovány pomocí rekurentních neuronových sítí, přináší efekt regularizace a umožňuje trénovat modely s větším počtem parametrů bez overfittingu. Na závěr popisujeme, které kombinace datasetů a nastavení modelů by mohly mít z adversálního tréninku největší prospěch.
[ "Petr Bělohlávek", "Ondřej Plátek", "Zdeněk Žabokrtský", "Milan Straka" ]
https://www.semanticscholar.org/paper/4e7470ff0a7f8885a1039ca6f7f95fe65b65f1a0/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
Results of the WMT18 Metrics Shared Task: Both characters and embeddings achieve good performance
Výsledky úlohy hodnocení kvality strojového překladu WMT17 Metrics Shared Task
This paper presents the results of the WMT18 Metrics Shared Task. We asked participants of this task to score the outputs of the MT systems in- volved in the WMT18 News Transla- tion Task with automatic metrics. We collected scores of 10 metrics and 8 re- search groups. In addition to that, we computed scores of 8 standard met- rics (BLEU, SentBLEU, chrF, NIST, WER, PER, TER and CDER) as base- lines. The collected scores were eval- uated in terms of system-level corre- lation (how well each metric’s scores correlate with WMT18 official man- ual ranking of systems) and in terms of segment-level correlation (how often a metric agrees with humans in judging the quality of a particular sentence rel- ative to alternate outputs). This year, we employ a single kind of manual eval- uation: direct assessment (DA).
Článek přináší výsledky soutěže ve vyhodnocování strojového překladu WMT18 Metrics Shared Task. Vyhodnocujeme 10 metrik získaných od 8 týmů a navíc 8 standardních metrik.
[ "Qingsong Ma", "Ondřej Bojar", "Yvette Graham" ]
https://www.semanticscholar.org/paper/b4e59a9926362fd18ce39c64bb42fe874b4de8f9/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
Lexicographic Potential of the Syntactic Properties of Verbs: The Case of Reciprocity in Czech
Lexikografický potenciál syntaktických vlastností sloves: Případy reciprocity v češtině
In this paper, we deal with reciprocity, its representation in a lexicon and its possible contribution to the description of verbs.
Příspěvěk se zabývá otázkou, jak může syntaktická vlastnost sloves, jako je reciprocita, být promítnuta do slovníka a přispět k dalšímu popisu a klasifikaci sloves.
[ "Václava Kettnerová", "Markéta Lopatková" ]
https://www.semanticscholar.org/paper/6360d031752905aeb1096f03061bc0305e62c8a1/
null
null
null
null
null
null
null
null
null
null
null
null
cs
2018
Between Reflexivity and Reciprocity: Towards an Analysis of Reflexive and Reciprocal Constructions of Selected Czech Verbs
Mezi reflexivitou a reciprocitou: Poznámky k reflexivním a recipročním konstrukcím vybraných českých sloves
In this article, we focus on two groups of Czech verbs, lexical and syntactic reciprocals. We provide an analysis of their syntactic properties with respect to reciprocity and reflexivity, their possible ambiguity and combination. We demonstrate that it is the reflexive pronoun and its surface expression that play a key role in the studied phenomena.
Na pozadí dvou skupin sloves, lexikálních a syntaktických reciprok v češtině, diskutujeme otázky homonymie a kombinovatelnosti reflexivity a reciprocity. Ukazujeme, že zdrojem homonymie i možnosti kombinace reflexivity a reciprocity je reflexivní zájmeno a že klíčovou úlohu sehrává jeho povrchověsyntaktická pozice.
[ "Václava Kettnerová", "Markéta Lopatková" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
Enriching VALLEX with Light Verbs: From Theory to Data and Back Again
Obohacení VALLEXu o kategoriální slovesa: Od teorie k datům a zpět
In this paper, we summarize a theoretical analysis of syntactic behavior of Czech light verb constructions and their verification in the data annotation of 1,500 Czech light verbs constructions, which have been integrated in VALLEX, version 3.5.
V tomto příspěvku shrnujeme teoretickou analýzu syntaktického chování českých komplexních predikátů s kategoriálním slovesem, která byla ověřena v anotaci 1500 komplexních predikátů. Tato anotace je součátí slovníku VALLEX, verze 3.5.
[ "Václava Kettnerová", "Markéta Lopatková", "Eduard Bejček", "Petra Barančíková" ]
https://www.semanticscholar.org/paper/1675a41d004f42277874904516a657778b9fcb24/
null
null
null
null
null
null
null
null
null
null
null
null
cs
2018
Towards Complement System in Contemporary Czech
Poznámky k systému tzv. komplementů v češtině
The talk dealt with the so-called complement system in Czech as it manifests in the data from the substitutability of dependent content clauses with infinitives.
Přednáška se zabývala tzv. komplementovým systémem v češtině, a to zejm. vztahem mezi závislými obsahovými klauzemi a infinitivy sloves.
[ "Václava Kettnerová", "Magda Ševčíková" ]
https://www.semanticscholar.org/paper/8958db8e5ddebd442fb7bcde7ea3a0def2c4a17d/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
The Forest Lion and the Bull: Morphosyntactic Annotation of the Panchatantra
Král lesů a buvol: Morfosyntaktická anotace Paňčatantry
We present the first freely available dependency treebank of Sanskrit. It is based on text from Panchatantra, an ancient Indian collection of fables. The annotation scheme we chose is that of Universal Dependencies, a current de-facto standard for cross-linguistically comparable morphological and syntactic annotation. In the present paper, we discuss word segmentation issues, morphological inventory and certain interesting syntactic constructions in the light of the Universal Dependencies guidelines. We also present an initial parsing experiment.
Popisujeme první volně dostupný závislostní korpus sanskrtu. Je založen na textech z Paňčatantry, starověké indické sbírky bajek. Zvolili jsme formalismus Universal Dependencies, který v současnosti představuje faktickou normu mezijazykově srovnatelné morfologické a syntaktické anotace. V článku probíráme obtíže se segmentací textu na slova, představujeme inventář morfologických kategorií, jakož i některé syntaktické konstrukce, které jsou zajímavé ve světle pravidel Universal Dependencies. Dále popisujeme experiment s automatickou syntaktickou analýzou (parsingem).
[ "Puneet Dwivedi", "Daniel Zeman" ]
https://www.semanticscholar.org/paper/9952cfa0457a96f385347ffe56ecd29c6f41eded/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
The PARSEME multilingual corpus of verbal multiword expressions
Mnohojazyčný korpus slovesných víceslovných výrazů PARSEME
Multiword expressions (MWEs) are known as a “pain in the neck” due to their idiosyncratic behaviour. While some categories of MWEs have been largely studied, verbal MWEs (VMWEs) such as to take a walk, to break one’s heart or to turn off have been relatively rarely modelled. We describe an initiative meant to bring about substantial progress in understanding, modelling and processing VMWEs. In this joint effort carried out within a European research network we elaborated a universal terminology and annotation methodology for VMWEs. Its main outcomes, available under open licenses, are unified annotation guidelines, and a corpus of over 5.4 million words and 62 thousand annotated VMWEs in 18 languages.
Článek popisuje mezinárodní projekt, který připravil univerzální terminologii a anotační postup pro slovesné víceslovné výrazy. Výstupem je korpus zveřejněný pod otevřenou licencí (18 jazyků, 5,4 milionů slov, 62 tisíc víceslovných výrazů) a univerzální anotační instrukce.
[ "Agata Savary", "Marie Candito", "Verginica Barbu Mititelu", "Eduard Bejček", "Fabienne Cap", "Slavomír Čéplö", "Silvio Ricardo Cordeiro", "Gülşen Cebiroğlu Eryiğit", "Voula Giouli", "Maarten van Gompel", "Yaakov Ha-Cohen Kerner", "Jolanta Kovalevskaitė", "Simon Krek", "Chaya Liebeskind", "Johanna Monti", "Carla Parra Escartín", "Lonneke van der Plas", "Behrang QasemiZadeh", "Carlos Ramisch", "Federico Sangati", "Ivelina Stoyanova", "Veronika Vincze" ]
https://www.semanticscholar.org/paper/9b48cbcaac4871a8264f9735ba323fe44c682951/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
CUNI NMT System for WAT 2018 Translation Tasks
UK překladový model pro WAT 2018
This paper describes the CUNI submission to WAT 2018 for the English-Hindi translation task using a transfer learning techniques which has proven effective under low resource conditions. We have used the Transformer model and utilized an English-Czech parallel corpus as additional data source. Our simple transfer learning approach first trains a “parent” model for a high-resource language pair (English-Czech) and then continues the training on the low-resource (English-Hindi) pair by replacing the training corpus. This setup improves the performance compared with the baseline and in combination with back-translation of Hindi monolingual data, it allowed us to win the English-Hindi task. The automatic scoring by BLEU did not correlate well with human judgments.
Tento článek popisuje CUNI system na WAT 2018 pro překlad mezi angličtinou a Hindštinou. Náš systém využívá transfer learningu z anglicko-českého modelu. Využíváme technologii neuronového transformeru. Náš systém začíná trénováním na jazykovém páru s mnoho paralelními daty, česko-anglickém, na který naváže jazykový pár s nedostatkem dat. Náš systém se umístil první podle lidského hodnocení.
[ "Tom Kocmi", "Shantipriya Parida", "Ondřej Bojar" ]
https://www.semanticscholar.org/paper/ad380a8dd7f7f7aa790f68acaf9478dc4c6d644d/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
Trivial Transfer Learning for Low-Resource Neural Machine Translation
Jednoduché přetrénování překladových modelů pro jazyky s nedostatkem dat
Transfer learning has been proven as an effective technique for neural machine translation under low-resource conditions. Existing methods require a common target language, language relatedness, or specific training tricks and regimes. We present a simple transfer learning method, where we first train a “parent” model for a high-resource language pair and then continue the training on a lowresource pair only by replacing the training corpus. This “child” model performs significantly better than the baseline trained for lowresource pair only. We are the first to show this for targeting different languages, and we observe the improvements even for unrelated languages with different alphabets.
Bylo dokázáno, že přetrénování natrénovaných modelů vede k zlepšení strojového překladu. Existující metody potřebují, aby zkoumané jazykové páry byly lingvisticky podobné. My představujeme přístup, který nepotřebuje lingvistickou podobnost jazyků ani specifické trénovací kroky.
[ "Tom Kocmi", "Ondřej Bojar" ]
https://www.semanticscholar.org/paper/6b8594a671878b64b11751b4af7a47c917253ccf/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
CUNI Submissions in WMT18
UK systém pro překlad novinových článků WMT 2018
We participated in the WMT 2018 shared news translation task in three language pairs: English-Estonian, English-Finnish, and English-Czech. Our main focus was the lowresource language pair of Estonian and English for which we utilized Finnish parallel data in a simple method. We first train a “parent model” for the high-resource language pair followed by adaptation on the related lowresource language pair. This approach brings a substantial performance boost over the baseline system trained only on Estonian-English parallel data. Our systems are based on the Transformer architecture. For the English to Czech translation, we have evaluated our last year models of hybrid phrase-based approach and neural machine translation mainly for comparison purposes.
Zúčastnili jsme se překladové soutěže WMT 2018 v překladu novinových článků. Zúčastnili jsme se v jazykových párech Angličtina-Estonština, Angličtina-Finština a Angličtina-Čeština. Náš hlavní cíl byly jazyky s nedostatkem trénovacích dat, tedy překlad mezi angličtinou a estonštinou.
[ "Tom Kocmi", "Roman Sudarikov", "Ondřej Bojar" ]
https://www.semanticscholar.org/paper/9809cb10cea5e91b74c489cc2c8f087f67e3d3b0/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
CUNI Basque-to-English Submission in IWSLT18
UK překladový model pro Baskičtinu
We present our submission to the IWSLT18 Low Resource task focused on the translation from Basque-to-English. Our submission is based on the current state-of-the-art self-attentive neural network architecture, Transformer. We further improve this strong baseline by exploiting available monolingual data using the back-translation technique. We also present further improvements gained by a transfer learning, a technique that trains a model using a high-resource language pair (Czech-English) and then fine-tunes the model using the target low-resource language pair (Basque-English).
Představujeme náš překladový systém pro překlad z Baskičtiny do Angličtiny. K jeho natrénování jsme využili metodu přetrénování jiného překladového modelu, v našem případě anglicko-českého.
[ "Tom Kocmi", "Dušan Variš", "Ondřej Bojar" ]
https://www.semanticscholar.org/paper/13c5d9128798a74c41da833e4fb672732a375803/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
Adding Words to Manuscripts: From PagesXML to TEITOK
Přidávání slov k rukopisům: Od PagesXML po TEITOK
This article describes a two-step method for transcribing historic manuscripts. In this method, the first step uses a page-based representation making it easy to transcribe the document page-by-page and line-by-line, while the second step converts this to the TEI/XML text-based format, in order to make sure the document becomes fully searchable.
Tento článek popisuje dvoustupňovou metodu přepisu historických rukopisů. V této metodě používá první krok reprezentaci na stránce, což usnadňuje přepis dokumentu po stránce a řádku po řádku, zatímco druhý krok to převádí do textového formátu TEI / XML, aby zajistit, aby byl dokument plně prohledávatelný.
[ "Maarten Janssen" ]
https://www.semanticscholar.org/paper/18bd4c461f537bb1ccb600081678011ff7529d2c/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
Quantitative Analysis of Syntactic Dependency in Czech
Kvantntativní analýza syntaktické závislosti v češtině
The article presents a quantitative analysis of some syntactic dependency properties in Czech. A dependency frame is introduced as a linguistic unit and its characteristics are investigated. In particular, a ranked frequencies of dependency frames are observed and modelled and a relationship between particular syntactic functions and the number of dependency frames is examined. For the analysis, the Czech Universal Dependency Treebank is used.
Kapitola prezentuje kvantitativní analýzu vybraných závislostních vztahů v češtině. Zavádíme pojem valenčního rámec jako lingvistické jednotky a zkoumáme jeho základní charakteristiky, zejména frekvenční charakteristiky a dále vztah mezi typem syntaktické funkce a množstvím různých valenčních rámců, jež tato funkce nese. Analýzu jsou priváděny na korpusu Czech Universal Dependency Treebank.
[ "Radek Čech", "Jiří Milička", "Ján Mačutek", "Michaela Koščová", "Markéta Lopatková" ]
https://www.semanticscholar.org/paper/04c379ce3f2f991461b549593dfa1930e89840c6/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
An interface between the Czech valency lexicon PDT-Vallex and corpus manager KonText
Rozhraní mezi českým valenčním slovníkem PDT-Vallex a správcem korpusů KonText
We present a user interface between the Czech valency lexicon, PDT-Vallex [1], and KonText1 – a web application for querying corpora available within the LINDAT/CLARIN project. KonText allows evaluation of simple and complex queries, displaying their results as concordance lines, computing frequency distribution, calculating association measures for collocations and further work with language data. For every verb in a concordance line, our interface allows to display information concerning its valency frame in a separate window if corresponding entries exist in PDT-Vallex, as well as a list of possible valency frames for that particular verb. Information concerning verb frame comprises verb lemma, frame elements with semantic roles, vocabularystyle description and examples from PDT-Vallex, Prague Dependency Treebank [2] and Prague Czech-English Dependency Treebank [3]. The information is requested by REST-API from the valency lexicon PDT-Vallex that contains over 11000 valency frames for more than 7000 verbs which occurred in Prague Dependency Treebank, Prague Czech-English Dependency Treebank or Prague Dependency Treebank of Spoken Czech. We use a fork of KonText application (developed by the Institute of the Czech National Corpus) that has been further extended by the Institute of Formal and Applied Linguistics to suit the needs of LINDAT/CLARIN project. The plugin we present provides a unique solution for Czech language that integrates valency information from the Czech valency lexicon with the means of querying Prague Dependency Treebank.
Představujeme uživatelské rozhraní mezi českým valenčním lexikonem, PDT-Vallex [1] a KonText1 - webová aplikace pro dotazování na korpusy dostupné v rámci projektu LINDAT / CLARIN. KonText umožňuje vyhodnocování jednoduchých a komplexních dotazů, zobrazování výsledků jako shodných linek, výpočet distribuce frekvence, výpočet asociačních opatření pro kolokace a dále práce s jazykovými daty. Pro každé sloveso ve shodné lince umožňuje naše rozhraní zobrazit informace o jeho valenčním rámci v samostatném okně, pokud existují odpovídající položky PDT-Vallex, stejně jako seznam možných valenčních rámců pro toto konkrétní sloveso. Informace týkající se slovesného rámu obsahuje lemma slovesa, prvky rámce se sémantickými rolami, slovník slovníku popis a příklady z PDT-Vallex, Praha Dependency Treebank [2] a Praha Česko-anglická závislostní stromová banka [3]. Informace jsou požadovány z REST-API z valenční lexikon PDT-Vallex, který obsahuje přes 11000 valenčních rámců pro více než 7000 sloves která se objevila v pražské Dependency Treebank, Praha Czech-English Dependency Treebank nebo Praha závislost Treebank mluveného češtiny. Používáme vidlici aplikace KonText (vyvinutá Ústavem českého národního korpusu), který byl dále rozšířen Ústavem formální a aplikované lingvistiky, aby vyhovovaly potřebám projektu LINDAT / CLARIN. Plugin jsme present poskytuje unikátní řešení pro český jazyk, které integruje valenční informace z Český valenční lexikon s prostředky dotazování na pražskou Závislostní bankou.
[ "Kira Droganova", "Eva Fučíková", "Anna Vernerová" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
Cross-Tagset Parsing Evaluation for Russian
Analýza parsování pro ruštinu
Cross-tagset parsing is based on the substitution of one annotation layer for another while processing data within one language. As often as not, either the native tagger or the dependency parser used in (pre-)annotation of the Gold treebank is not available. The cross-tagset approach allows one to annotate new texts using freely available tools or tools optimized to user's needs. We evaluate the robustness of Russian dependency parsing using different morphological and syntactic tagsets in input and output. A qualitative analysis of errors shows that the cross-substitution of three morphological tagsets and two syntactic tagsets causes only a mild drop in performance.
Rozbalení pomocí křížových značek je založeno na nahrazení jedné vrstvy anotací za jinou při zpracování dat v jednom jazyce. Nejčastěji není k dispozici ani rodný značkovač nebo syntaktický analyzátor závislostí používaný v (před) anotaci Gold stromové banky. Přístup přes křížové štítky umožňuje anotovat nové texty pomocí volně dostupných nástrojů nebo nástrojů optimalizovaných podle potřeb uživatele. Vyhodnocujeme robustnost analyzování ruské závislostí s použitím různých morfologických a syntaktických tagů ve vstupu a výstupu. Kvalitativní analýza chyb ukazuje, že křížová substituce tří morfologických značek a dvou syntaktických značek způsobuje pouze mírný pokles výkonu.
[ "Kira Droganova", "Olga Lyashevskaya" ]
https://www.semanticscholar.org/paper/7a940e35dd4bf615af113c393252683f2df86acb/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
Data Conversion and Consistency of Monolingual Corpora: Russian UD Treebanks
Převod dat a konzistence jednojazyčných korpusů: ruské treebanky UD
In this paper we focus on syntactic annotation consistency within Universal Dependencies (UD) treebanks for Russian: UD_Russian-SynTagRus, UD_Russian-GSD, UD_Russian-Taiga, and UD_Russian-PUD. We describe the four treebanks, their distinctive features and development. In order to test and improve consistency within the treebanks, we reconsidered the experiments by Martínez Alonso and Zeman; our parsing experiments were conducted using a state-of-the-art parser that took part in the CoNLL 2017 Shared Task. We analyze error classes in functional and content relations and discuss a method to separate the errors induced by annotation inconsistency and those caused by syntactic complexity and other factors.
V tomto článku se zaměřujeme na syntaktickou anotaci a její konzistenci v korpusech Universal Dependencies (UD) pro ruštinu: SynTagRus, GSD, Taiga a PUD. Popisujeme tyto čtyři korpusy, jejich distinktivní rysy a vývoj.
[ "Kira Droganova", "Olga Lyashevskaya", "Daniel Zeman" ]
https://www.semanticscholar.org/paper/e64651f798d78fc76b17fca5be17aa832890b507/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
Artificial Treebank with Ellipsis
Umělý treebank s elipsou
Artificial dependency trees in the Universal Dependencies v2 style, focused on gapping (the 'orphan' relation in UD). For motivation and description of the data, see Droganova et al., 2018 (LREC, Miyazaki, Japan).
Umělé závislostní stromy v anotačním stylu Universal Dependencies v2, zaměřené na druh elipsy zvaný anglicky gapping (v UD odpovídá vztahu 'orphan'). Motivace a popis těchto dat je obsažen v Droganova et al., 2018 (LREC, Miyazaki, Japonsko).
[ "Kira Droganova", "Daniel Zeman", "Jenna Kanerva", "Filip Ginter" ]
https://www.semanticscholar.org/paper/6b5a5fc8d37c8fd5982fab50afdb149422811404/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
Parse Me if You Can: Artificial Treebanks for Parsing Experiments on Elliptical Constructions
Rozeber mě, když to dokážeš: Umělé korpusy pro pokusy s automatickou syntaktickou analýzou eliptických konstrukcí
In this work we focus on a particular linguistic phenomenon, ellipsis, and explore the latest parsers in order to learn about parsing accuracy and typical errors from the perspective of elliptical constructions. For this purpose we collected and processed outputs of several state-of-the art parsers that took part in the CoNLL 2017 Shared Task. We extended the official shared task evaluation software to obtain focused evaluation of elliptical constructions. Since the studied structures are comparatively rare, and consequently there is not enough data for experimentation, we further describe the creation of a new resource, a semi-artificially constructed treebank of ellipsis.
V této práci se zaměřujeme na konkrétní jazykový jev, elipsu, a zkoumáme výstupy současných parserů, abychom zjistili jejich úspěšnost a typické chyby s ohledem na elipsy. K tomuto účelu jsme sebrali a zpracovali výstupy několika nejlepších parserů, které se zúčastnily společné úlohy CoNLL 2017. Oficiální vyhodnocovací software jsme rozšířili, aby bylo možné zjistit chyby v analýze elips. Protože studované struktury jsou poměrně vzácné, a tudíž není k dispozici dostatečné množství dat pro experimenty, popisujeme dále tvorbu nového datového zdroje, polouměle vytvořeného závislostního korpusu elips.
[ "Kira Droganova", "Daniel Zeman", "Jenna Kanerva", "Filip Ginter" ]
https://www.semanticscholar.org/paper/d8ca19e56a337706d33b64963c851857bf1e5736/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
Mind the Gap: Data Enrichment in Dependency Parsing of Elliptical Constructions
Pozor na díry: Obohacování dat pro závislostní analýzu eliptických konstrukcí
We report on experiments with several approaches to automatically extending training data for dependency parsers, using large crawled web corpora. One set of methods is general, draws upon self-training and tri-training and adds a novel algorithm of mimicking the structural complexity of the original treebank. Methods from the other set are more focused on elliptical constructions. We provide evaluation on 5 languages: Czech, English, Finnish, Russian and Slovak.
Popisujeme pokusy s několika přístupy k automatickému rozšíření trénovacích dat pro závislostní syntaktické analyzátory s využitím velkých webových korpusů. Jedna sada metod je obecná, inspiruje se samotrénováním a trojtrénováním a přidává nový algoritmus, který napodobuje strukturální složitost původního treebanku. Metody ve druhé sadě se více zaměřují na eliptické konstrukce. Pokusy vyhodnocujeme na 5 jazycích: češtině, angličtině, finštině, ruštině a slovenštině.
[ "Kira Droganova", "Filip Ginter", "Jenna Kanerva", "Daniel Zeman" ]
https://www.semanticscholar.org/paper/908859cee19563d703c188751e202c8810223113/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
Discussion Group Summary: Optical Music Recognition
Shrnutí diskusní skupiny: Rozpoznávání notopisu
This document summarizes the discussion of the interest group on Optical Music Recognition (OMR) that took place in the 12th IAPR International Workshop on Graphics Recognition, and presents the main conclusions drawn during the session: OMR should revisit how it describes itself, and the OMR community should intensify its collaboration both internally and with other stakeholders.
Tento příspěvek shrnuje diskusi skupiny aktivních výzkumníků rozpoznávání notopisu (Optical Music Recognition, OMR), která se konala v rámci 12th IAPR International Workshop on Graphics Recognition, a prezentuje její výstupy: OMR by mělo zpřesnit svou terminologickou a taxonomickou základnu, a komunita výzkumníků v oboru by měla intenzivněji spolupracovat jak mezi sebou, tak se zájemci o OMR.
[ "Jorge Calvo-Zaragoza", "Jan Hajič, jr.", "Alexander Pacha" ]
https://www.semanticscholar.org/paper/63efea6356750c555e2893b4b54e6616625e650b/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
LemmaTag: Jointly Tagging and Lemmatizing for Morphologically Rich Languages with BRNNs
LemmaTag: Společné značkování a lemmatizace morfologicky bohatých jazyků pomocí BRNN
We present LemmaTag, a featureless neural network architecture that jointly generates part-of-speech tags and lemmas for sentences by using bidirectional RNNs with character-level and word-level embeddings. We demonstrate that both tasks benefit from sharing the encoding part of the network, predicting tag subcategories, and using the tagger output as an input to the lemmatizer. We evaluate our model across several languages with complex morphology, which surpasses state-of-the-art accuracy in both part-of-speech tagging and lemmatization in Czech, German, and Arabic.
Představujeme LemmaTag, architekturu neuronové sítě, která společně generuje morfologické značky a lemmata pomocí obousměrných rekurentních neuronových sítí pomocí slovních a znakových embeddingů. Demonstrujeme, že oběma úkolům pomáhá sdílet enkodér, předvídat podtypy značek a používat předpovězené značky na vstup lemmatizátoru. Vyhodnocujeme náš model na několika jazycích se složitou morfologií, a překonáváme nejlepší známé výsledky jak morfologického značkování tak lemmatizace v češtině, němčině a arabštině.
[ "Daniel Kondratyuk", "Tomáš Gavenčiak", "Milan Straka", "Jan Hajič" ]
https://www.semanticscholar.org/paper/2e531d04bb8b64a7a76530d3e5f272d494d84812/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
Collective memory as an interactional practice: The case of the Czech-Jewish experience in Switzerland during the WWII period
Kolektivní paměť jako interakční praxe: Případ česko-židovské zkušenosti ve Švýcarsku v době druhé světové války
In 2017/2018, I was working on a project „Collective memory as an interactional practice: The case of the Czech-Jewish experience in Switzerland during the World War II period”, carried out at the University of Fribourg in Switzerland. My research consisted of three successive stages with mutually interrelated aims: (1) Analysis of oral history interviews with Holocaust survivors born in the former Czechoslovakia, who have spent time in Switzerland as refugees, asylum seekers or displaced persons during the World War II period. The oral history interviews were selected from the USC Shoah Foundation’s Visual History Archive. (2) Short clips of interviews were then incorporated into on-line educational material, accompanied by text and images. The resulting on-line lesson had the following structure: (I) The context (Why did people leave Czechoslovakia?); (II) The decision (How do the narrators reflect the decision to leave Czechoslovakia?); (III) The journey (How was the emigration practically accomplished?); (IV) The memory: (How is the migration retrospectively reflected by the narrators?). (3) The on-line lesson was tested at five schools in Switzerland and in the Czech Republic between January and March 2018. The sessions were videotaped and analysed in detail: not only to evaluate and improve the developed activities, but also to explore the interactional specifics of the educational setting. The research project provides an example of using oral history interviews as cases of narrative expression of the refugee experience during the World War II period. It explored one of the possible ways of utilizing large archives of oral histories for the transmission of historical knowledge on refugeedom, focusing on personal aspects. Moreover, it highlights several issues in educational use of archival interviews, such as the necessity of sufficient context for the students’ comprehension, and also the more general topics of incorporating digital technologies into classroom interaction.
Prezentace výzkumného projektu a jeho výsledků je v kontextu workshopu příkladem využití sbírek orálněhistorických interview jako příkladů narativního vyjádření uprchlické zkušenosti během druhé světové války. Důraz je kladen na vzdělávací uplatnění úryvků z rozhovorů prostřednictvím informačních technologií ve výuce.
[ "Jakub Mlynář" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
cs
2018
Practical temporality of work with digital oral history at three Czech high schools
Praktická časovost práce s digitální orální historií na třech gymnáziích v ČR
The paper is based on analysis of situated action in testing lessons, which were implemented in the school year 2017/18 at three high schools in the Czech Republic. Video recordings taken over the course of the lessons were then subjected to a detailed analysis grounded in the tradition of ethnomethodology and conversation analysis.
Příspěvek se zaměřil na analýzu situovaného jednání v rámci testovacích vyučovacích hodin, jež byly realizovány ve školním roce 2017/18 na třech gymnáziích v ČR. Během těchto hodin studenti pracovali v menších skupinách 2–3 osob s on-line materiálem složeným z textů, obrázků a klipů vyňatých z audiovizuálních orálněhistorických rozhovorů. Videonahrávky pořízené v průběhu hodin byly následně podrobeny detailní analýze vycházející z tradice etnometodologie a konverzační analýzy. V prezentaci vybraných výsledků analýzy se soustředím především na specifika začlenění videoklipu do sociální interakce. Samotná práce s videoklipem je účastníky metodicky zasazena do širšího kontextu dalších aktivit, zvláště četby úvodních textů na monitoru a následného zapisování odpovědí do pracovního listu. Z hlediska sekvenčnosti práce s videoklipem lze rozlišit tři fáze, a to (i) přípravu na spuštění videa, během níž je třeba dosáhnout optimálního uspořádání materiálních artefaktů a tělesné orientace účastníků; (ii) sledování videa, kde je zřejmá preference nepřerušeného přehrávání od začátku do konce, stejně jako omezení řečových projevů účastníků na „průběžný komentář“; (iii) reflexe videa, která nejčastěji směřuje k rychlé orientaci na formulaci odpovědi a její zápis do pracovního listu. Ukážu, že videoklip je v analyzovaných interakcích přítomen jako zdroj (např. dočasné pozastavení klipu či jeho opakované sledování za účelem identifikace klíčových pasáží) i jako téma (např. komentáře týkající se délky klipu či jazyka mluvčího, jenž na nahrávce hovoří). Příspěvek tak skrze rozbor konkrétních událostí vypovídá i o roli moderních digitálních technologií v současné společnosti a ve vzdělávání.
[ "Jakub Mlynář" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
cs
2018
"We all had the feeling -- we have to get rid of it, we have to forget": Reflection of Holocaust commemoration in interviews with Czech and Slovak survivors
„To jsme všechny měly ten pocit – musíme se toho zbavit, musíme na to zapomenout“: Reflexe připomínání holocaustu v rozhovorech s česko-slovenskými přeživšími
This chapter presents a conceptual framework for approaching the topic of Holocaust commemoration, based on the analysis of selected interviews from the USC Shoah Foundation's Visual History Archive.
V kapitole je předložen základní rámec pro uchopení tématu připomínání holocaustu, založený na analýze vybraných interview z Archivu vizuální historie USC Shoah Foundation.
[ "Jakub Mlynář" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
Towards a Sociological Conception of Artificial Intelligence
K sociologickému pojetí umělé inteligence
Social sciences have been always formed and influenced by the development of society, adjusting the conceptual, methodological, and theoretical frameworks to emerging social phenomena. In recent years, with the leap in the advancement of Artificial Intelligence (AI) and the proliferation of its everyday applications, “non-human intelligent actors” are increasingly becoming part of the society. This is manifested in the evolving realms of smart home systems, autonomous vehicles, chatbots, intelligent public displays, etc. In this paper, we present a prospective research project that takes one of the pioneering steps towards establishing a “distinctively sociological” conception of AI. Its first objective is to extract the existing conceptions of AI as perceived by its technological developers and (possibly differently) by its users. In the second part, capitalizing on a set of interviews with experts from social science domains, we will explore the new imaginable conceptions of AI that do not originate from its technological possibilities but rather from societal necessities. The current formal ways of defining AI are grounded in the technological possibilities, namely machine learning methods and neural network models. But what exactly is AI as a social phenomenon, which may act on its own, can be blamed responsible for ethically problematic behavior, or even endanger people’s employment? We argue that such conceptual investigation is a crucial step for further empirical studies of phenomena related to AI’s position in current societies, but also will open up ways for critiques of new technological advancements with social consequences in mind from the outset.
Text pojednává o možných přístupech ke zkoumání umělé inteligence jako sociálního fenoménu.
[ "Jakub Mlynář", "Hamed Alavi", "Himanshu Verma", "Lorenzo Cantoni" ]
https://www.semanticscholar.org/paper/64936695ed63a2a3b1170f81f66b4b2be9f2d3db/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
Translation of “and” in a parallel TED Talk corpus of English, Czech, Hungarian, Lithuanian and French: functions and omissions
Překlady "and" na materiálu paralelního korpusu Ted talks v angličtině, češtině, maďarštině, litevštině a francouzštině: funkce a vynechávky
In this paper we report on the methods and findings of a multilingual corpus study focusing on the functions of and in English and its translations into Czech, French, Hungarian and Lithuanian, in a selection of TedTalks.
V tomto příspěvku popisujeme metody a výsledky výzkumu zaměřeného na vícejazyčné korpusové srovnání spojovacího výrazu "and" v angličtině a jeho ekvivalentů v češtině, francouzštině, maďarštině a litevštině u vybraných příspěvků z TED talks.
[ "Ágnes Abuczki", "Nijolė Burkšaitienė", "Ludivine Crible", "Anna Nedoluzhko", "Péter Furkó", "Giedre Valunaite Oleskeviciene", "Sigita Rackevičienė", "Šárka Zikánová" ]
https://www.semanticscholar.org/paper/c0a3a5a43d84a51adc0b86ddc14c7d2ebf4238a3/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
The underspecified connective "and" in a parallel TED Talk corpus: functions, translation and implicitation
Podspecifikovaný konektor "and" v paralelním korpusu TED talks: funkce, překlad a nevyjádřenost
Analysis of functions of the discourse connective AND and its counterparts in English, Czech, Lithuanian, French and Hungarian according to the taxonomy with cross-domain functions (Crible, Degand, 2017).
Analýza funkcí spojovacího výrazu AND a jeho ekvivalentů v angličtině, češtině, litevštině, francouzštině a maďarštině podle doménové klasifikace Crible - Degand (2017).
[ "Ágnes Abuczki", "Nijolė Burkšaitienė", "Ludivine Crible", "Péter Furkó", "Anna Nedoluzhko", "Giedre Valunaite Oleskeviciene", "Šárka Zikánová" ]
https://www.semanticscholar.org/paper/54e3945997990c493f4ecc448ccb1b473dbbec3f/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
Translation of "and" in a parallel TED Talk corpus of English, Czech, Hungarian, Lithuanian and French: functions and omissions
Překlad spojky "a" v paralelním korpusu TED Talk corpus v angličtině, češtině, maďarštině, litevštině a francouzštině: funkce a výpustky
Cross-linguistic comparison of the functions of the conjunction "and" and the possibilities of its translation, based on the parallel translations of subtitles in TED talks. Pilot study about the underspecification of semantics of some conjunctions in different languages.
Mezijazykové srovnání funkcí spojky "a" a možností jejího překladu, založené na paralelních překladech titulků v TED talks. Výchozí studie o nespecifičnosti významu některých spojek v různých jazycích.
[ "Ágnes Abuczki", "Nijolė Burkšaitienė", "Ludivine Crible", "Péter Furkó", "Giedre Valunaite Oleskeviciene", "Sigita Rackevičienė", "Šárka Zikánová" ]
https://www.semanticscholar.org/paper/c0a3a5a43d84a51adc0b86ddc14c7d2ebf4238a3/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
Morphological segmentation of Czech Words
Morfologická segmentace českých slov
In linguistics, words are usually considered to be composed of morphemes: units that carry meaning and are not further subdivisible. The task of this thesis is to create an automatic method for segmenting Czech words into morphemes, usable within the network of Czech derivational relations DeriNet. We created two different methods. The first one finds morpheme boundaries by differentiating words against their derivational parents, and transitively against their whole derivational family. It explicitly models morphophonological alternations and finds the best boundaries using maximum likelihood estimation. At worst, the results are slightly worse than the state of the art method Morfessor FlatCat, and they are significantly better in some settings. The second method is a neural network made to jointly predict segmentation and derivational parents, trained using the output of the first method and the derivational pairs from DeriNet. Our hypothesis that such joint training would increase the quality of the segmentation over training purely on the segmentation task seems to hold in some cases, but not in other. The neural model performs worse than the first one, possibly due to being trained on data which already contains some errors, multiplying them.
V lingvistice se obvykle slova považují za složená z morfémů, což jsou dále nedělitelné jazykové jednotky nesoucí význam. Zadáním této práce je nalézt automatickou metodu dělení českých slov na morfémy, které by bylo možné přidat do DeriNetu, sítě derivačních vztahů mezi českými slovy. Vytvořili jsme dvě různé takové metody. První nalézá hranice morfémů na základě hledání rozdílů mezi slovem a jeho derivačním předkem, a tranzitivně mezi všemi slovy v derivačním hnízdě. Tato metoda explicitně modeluje hláskové a morfologické alternace a nalézá nejvhodnější hranice morfémů pomocí metody maximální věrohodnosti. Ve srovnání s moderním systémem Morfessor FlatCat naše metoda přinejhorším mírně zaostává, ovšem v některých testech naopak dosahuje výsledků výrazně lepších. Druhou metodou je neuronová síť pro současné předpovídání morfologické segmentace a derivačních předků, trénovaná na datech získaných první metodou a na derivačních vztazích ze sítě DeriNet. S naší hypotézou, že tento způsob trénování dvou úloh naráz pomůže k dosažení lepších výsledků oproti trénování samotné segmentace, jsou však ve shodě pouze některé provedené pokusy. Celkově dosahuje neuronová síť horších výsledků než první metoda, pravděpodobně kvůli trénování na datech obsahujících chyby, které se tím přidávají k chybám metody samotné.
[ "Jonáš Vidra" ]
https://www.semanticscholar.org/paper/264a3d29d9aafd2d438c403c4827cbba0533457d/
null
null
null
null
null
null
null
null
null
null
null
null
cs
2018
DeriNet version 1.6
DeriNet verze 1.6
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 derivational relations between a derived word and its base word. The present version, DeriNet 1.6, contains 1,027,832 lexemes (sampled from the MorfFlex dictionary) connected by 803,404 derivational links. Furthermore, starting with version 1.5, DeriNet contains annotations related to compounding (compound words are distinguished by a special mark in their part-of-speech labels). Compared to version 1.5, version 1.6 was expanded by extracting potential links from dictionaries available under suitable licences, such as Wiktionary, and by enlarging the number of marked compounds.
DeriNet je lexikální síť modelující derivační vztahy v češtině. Uzly odpovídají lexémům, hrany odpovídají slovotvorným derivacím. Současná verze, DeriNet 1.6, obsahuje 1 027 832 lexémů převzatých ze slovníku MorfFlex, spojených 803 404 derivačními vztahy. Kromě toho obsahuje DeriNet, počínaje verzí 1.5, anotace kompozit (kompozita jsou označena vyhrazeným znakem ve svém záznamu o slovním druhu). Oproti verzi 1.5 byla verze 1.6 rozšířena o vztahy extrahované ze slovníků dostupných pod svobodnými licencemi, například Wiktionary, a o množství kompozitních značek.
[ "Jonáš Vidra", "Lukáš Kyjánek", "Magda Ševčíková", "Zdeněk Žabokrtský", "Adéla Kalužová", "Šárka Dohnalová", "Vojtěch Hudeček" ]
https://www.semanticscholar.org/paper/adacaa5bf522c85753cf0d3451712dc3f520e20d/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
Implementation of an Open Science Policy in the context of management of CLARIN language resources: a need for changes?
Implementace pravidel Open Science v kontextu správy jazykových dat v CLARINu: potřeba změn?
This article investigates the compatibility of the current CLARIN license categorization scheme with the open science paradigm. The first part presents the main concepts and theoretical framework required for the analysis, while the second part discusses the use of the CLARIN categorization system, divided into PUB (public), ACA (academic), and RES (restricted), and potential ways to change it. This paper serves to explore various suggestions for change and to begin discussion of a reformed CLARIN license category scheme.
Článek analyzuje kompatibilitu stávající kategorizace licencí v CLARINu s paradigmatem Open Science. V první části se prezentují základní koncepty a teoretický rámec, druhá část řeší kategorizaci licencí v projektu CLARIN do tříd PUB, ACA a RES a možnosti do budoucna tento systém změnit. Varianty možných změn jsou analyzovány jako podklad k další diskusi o změně přístupu k licencím v CLARINu.
[ "Aleksei Kelli", "Krister Lindén", "Kadri Vider", "Penny Labropoulou", "Erik Ketzan", "Paweł Kamocki", "Pavel Straňák" ]
https://www.semanticscholar.org/paper/c2a27129681b040fac3931c8236148ee387539f4/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
Reflexives in Universal Dependencies
Reflexiva v Universal Dependencies
We explore the annotation of reflexives in Universal Dependencies (UD) 2.2 treebanks (Nivre et al., 2018), with a stronger focus on Slavic languages. We have tried to find out if the current guidelines are transparent and clear enough for the annotators to follow them successfully. We point out a number of inconsistencies in the current annotation across languages, and propose improvements—sometimes of the guidelines, but mostly of the annotation. The goal of the paper is to contribute to more consistent and cross-linguistically parallel annotation of reflexives in the future releases of UD.
Zkoumáme anotaci zvratných zájmen v korpusech Universal Dependencies (UD) 2.2 (Nivre et al., 2018), přičemž se zaměřujeme zejména na slovanské jazyky. Snažíme se zjistit, zda jsou současná anotační pravidla dostatečně jasná, aby se jimi anotátoři dokázali řídit. Ukazujeme celou řadu nesrovnalostí napříč jazyky v současných datech a navrhujeme vylepšení–někdy anotačních pravidel, většinou však přímo anotace dat. Cílem článku je přispět ke konzistentnější a mezijazykově paralelnější anotaci reflexiv v budoucích vydáních UD.
[ "Sonja Marković", "Daniel Zeman" ]
https://www.semanticscholar.org/paper/b5e65d409386ac5177f95968bbd02c71e95708f6/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
Analysis of coreferential expressions in PAWS
Analyza koreferencnich vztahu v PAWS
In this paper, we decribe the coreference annotation on a multi-lingual parallel treebank (PAWS), a portion of Wall Street Journal translated into Czech, Russian and Polish which continues the tradition of multilingual treebanks with coreference annotation. The paper focuses on language-specific differences. We analyse syntactic structures concerning anaphoric relations in the languages under analysis, such as personal and impersonal constructions in polypredicative constructions and pro-drop qualities.
V článku popisujeme anotaci koreferenčních vztahů v multijazykovém paralelním korpusu PAWS. Zaměřujeme se na mezijazykové rozdíly ve vejadřování koreference, jako např. osobní a neosobní slovesné tvary v polypredikativních konstrukcích.
[ "Anna Nedoluzhko", "Michal Novák", "Maciej Ogrodniczuk" ]
https://www.semanticscholar.org/paper/fda43aa3c1d21d3f1673eeef60204ec12b1ce6e7/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
PAWS
PAWS
PAWS (Parallel Anaphoric Wall Street Journal) is a multi-lingual parallel treebank with coreference annotation. It consists of English texts from the Wall Street Journal translated into Czech, Russian and Polish. In addition, the texts are syntactically parsed and word-aligned. PAWS is based on PCEDT 2.0 and continues the tradition of multilingual treebanks with coreference annotation. PAWS offers linguistic material that can be further leveraged in cross-lingual studies, especially on coreference.
PAWS je anglicko-česko-rusko-polský korpus s anotací koreference.
[ "Anna Nedoluzhko", "Michal Novák", "Maciej Ogrodniczuk" ]
https://www.semanticscholar.org/paper/fc09d6486be1c9bbfbef4165ce3c1ab664e5d084/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
PAWS: A Multi-lingual Parallel Treebank with Anaphoric Relations
PAWS: Vícejazykový paralelní treebank s anotací anaforických vztahů
We present PAWS, a multi-lingual parallel treebank with coreference annotation. It consists of English texts from the Wall Street Journal translated into Czech, Russian and Polish. In addition, the texts are syntactically parsed and word-aligned. PAWS is based on PCEDT 2.0 and continues the tradition of multilingual treebanks with coreference annotation. The paper focuses on the coreference annotation in PAWS and its language-specific differences. PAWS offers linguistic material that can be further leveraged in cross-lingual studies, especially on coreference.
V článku představujeme PAWS, vícejazykový paralelní treebank s anotací koreference.
[ "Anna Nedoluzhko", "Michal Novák", "Maciej Ogrodniczuk" ]
https://www.semanticscholar.org/paper/7c1b45c37b2297a4f7ce5561caa39ae667d47b24/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
Correlating DRDs with other types of discourse phenomena: Cross-linguistic analysis of the interplay between DRDs, coreference and bridging
Vztah DRD s jinými typy diskurzních prostředků z kontrastivního pohledu
Since translations are influenced by various factors of translation process, it is often difficult to explain the real reasons of a certain construction used in translation data. The present contribution describes a cross-linguistic analysis of the interplay between discourse-relational devices (DRDs) and other discourse-related phenomena, such as coreference and bridging relations. The difference between the phenomena under analysis lies in the type of relations, which is expressed by a corresponding device. DRDs express logico-semantic relations between propositions, such as contrast, time, addition and others). Coreference serves the task of linking identical objects or events (i.e. complex anaphors) and bridging anaphora expresses non-identical or near-identical relations between referents, linking them with semantic interconnection.
Provádíme analýzu textových prostředků (na angličtině, češtině a němčitě), která je založená na textech, přeložených z angličtiny. Zvláštní důraz této publikace klademe na vlivu překladového faktoru na použití různých typů koheze.
[ "Anna Nedoluzhko", "Ekaterina Lapshinova-Koltunski" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
Interplay between conjunctions, coreference and bridging in Geman, English and Czech
Souhra konektorů, koreference a asociační anafory v němčině, angličtině a češtině
The purpose of our study is to analyse the interplay between conjunctions and other discourse-related phenomena, such as coreference and bridging relations in German, English and Czech. The difference between the phenomena under analysis lies in the type of relations, which is expressed by a corresponding device.
Cílem výzkumu je srovnání souhry konektorů, koreference a asociační anafory ve němčině, češtině a angličtině.
[ "Anna Nedoluzhko", "Ekaterina Lapshinova-Koltunski" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
Pronominal Adverbs in German and their Equivalents in English, Czech and Russian: Evidence from the Parallel Corpus
Pronominální advérbia v němčině a jejich ekvivalenty v angličtině, češtině a ruštině
The paper presents a contrastive analysis of pronominal adverbs in German (dabei, darauf, damit etc.) and their equivalents in English, Czech and Russian. The analysis is based on an empirical study of parallel news texts. Our main focus is to show the interplay between cohesive devices expressed through German pronominal adverbs in text and explore their equivalents in English, Czech and Russian. As the dataset at hand contains translations, we also focus on the influence of the translation factor in parallel texts.
V článku představíme kontrastivní analýzu zájmenných adverbií v němčině (dabei, daauf, damit aj.) a jejich ekvivalentů v angličtině, češtině a rušině. Analýza je založena na empirickém výzkumu paralelních textů.
[ "Anna Nedoluzhko", "Ekaterina Lapshinova-Koltunski" ]
https://www.semanticscholar.org/paper/4c224ca7c1a51bdcf34c836e9083de18fa9aa77d/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
Learning Audio–Sheet Music Correspondences for Cross-Modal Retrieval and Piece Identification
Učení korespondencí mezi zvukem a notovým záznamem pro vyhledávání napříč modalitami
This work addresses the problem of matching musical audio directly to sheet music, without any higher-level abstract representation. We propose a method that learns joint embedding spaces for short excerpts of audio and their respective counterparts in sheet music images, using multimodal convolutional neural networks. Given the learned representations, we show how to utilize them for two sheet-music-related tasks: (1) piece/score identification from audio queries and (2) retrieving relevant performances given a score as a search query. All retrieval models are trained and evaluated on a new, large scale multimodal audio–sheet music dataset which is made publicly available along with this article. The dataset comprises 479 precisely annotated solo piano pieces by 53 composers, for a total of 1,129 pages of music and about 15 hours of aligned audio, which was synthesized from these scores. Going beyond this synthetic training data, we carry out first retrieval experiments using scans of real sheet music of high complexity (e.g., nearly the complete solo piano works by Frederic Chopin) and commercial recordings by famous concert pianists. Our results suggest that the proposed method, in combination with the large-scale dataset, yields retrieval models that successfully generalize to data way beyond the synthetic training data used for model building.
Tento článek se věnuje přiřazování audia přímo k notopisu reprezentovanému jako obraz, bez jakýchkoliv abstratktních reprezentací. Navrhujeme metodu, která se naučí společný prostor pro reprezentaci krátkých útržků audia a jejich protějšků v obrázkách not pomocí multimodálních konvolučních neuronových sítí. Následně ukazujeme, jak s těmito naučenými reprezentacemi (1) identifikovat příslušnou skladbu podle nahrávky, (2) vyhledávat nahrávky pomocí obrázků not. Všechny vyhledávací modely jsou natrénované na novém velkém multimodálním datasetu audia a notopisu, který je spolu s tímto článkem dán veřejně k dispozici. Dataset obsahuje 479 detailně anotovaných klavírních skladeb od 53 skladatelů, celkem 1129 stran not a více než 15 hodin k nim zarovnaného audia, které bylo z příslušných not syntetizováno. Nad modelem natrénovaným těmito syntetickými daty však provádíme pokusy, které vyhledávají v databázi komplexních not (např. téměř celé dílo pro sólový klavír F. Chopina) a komerčních nahrávek špičkových klavíristů. Naše výsledky naznačují, že navržená metoda spolu s velkým datasetem vede k vyhledávacím modelům, které úspěšně zobecňují ze syntetických trénovacích dat na skutečná data.
[ "Matthias Dorfer", "Jan Hajič, jr.", "Andreas Arzt", "Harald Frostel", "Gerhard Widmer" ]
https://www.semanticscholar.org/paper/f1793b483a464e41b18f4d845d272599bd330020/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
Attention as a Perspective for Learning Tempo-invariant Audio Queries
Mechanismus pozornosti pro vyhledávání nahrávkou invariantní vůči tempu
Current models for audio–sheet music retrieval via multimodal embedding space learning use con- volutional neural networks with a fixed-size win- dow for the input audio. Depending on the tempo of a query performance, this window captures more or less musical content, while notehead den- sity in the score is largely tempo-independent. In this work we address this disparity with a soft attention mechanism, which allows the model to encode only those parts of an audio excerpt that are most relevant with respect to efficient query codes. Empirical results on classical piano music indicate that attention is beneficial for retrieval performance, and exhibits intuitively appealing behavior.
Současné modely pro křížové vyhledávání mezi nahrávkami a notovými zápisy přes natrénované multimodální reprezentace používají konvoluční neuronové sítě, jenž očekávají na vstupu pro audio pevně daný časový úsek. Podle tempa nahrávky však toto okno zachycuje různá množství hudebních událostí (not), zatímco fixně dlouhé okno do not jich zachycuje množství, které na tempu téměř nezávisí. V této práci se snažíme tento problém obejít pomocí mechanismu měkké pozornosti, která umožňuje modelu zakódovat pouze ty části útržku audia, které jsou nejvíce relevantní pro vybudování efektivního vyhledávacího klíče. Experimentální výsledky na klasické klavírní hudbě ukazují, že pozornostní mechanismus vyhledávání zlepšuje, a má intuitivně jasné výhody.
[ "Matthias Dorfer", "Jan Hajič, jr.", "Gerhard Widmer" ]
https://www.semanticscholar.org/paper/e76b268cb955b67441e6201c882ff0fe2caa736b/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
Paraphrases of verbal multiword expressions: the case of Czech light verbs and idioms
Parafráze slovesných víceslovných výrazů: Případ českých kategoriálních sloves a idiomů
In this chapter, we explore paraphrasability of Czech verbal MWEs (light verbs constructions and idioms) by single verbs in a semiautomatic experiment using word embeddings. Further, we propose a lexicographic representation and demonstrate one of its practical application in a machine translation experiment.
Tento příspěvek se zabývá parafrázovatelností českých slovesných víceslovných výrazů (kategoriálních slovesa a idiomů). V příspěvku je navržena lexikografická reprezentace parafrází a demonstrováno jejich praktické využití ve strojovém překladu.
[ "Petra Barančíková", "Václava Kettnerová" ]
https://www.semanticscholar.org/paper/312c2ed7a88e626bf1f32efcbe0678d232127851/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
CUNI team: CLEF eHealth Consumer Health Search Task 2018
CUNI v CLEF eHealth Consumer Health Search Task 2018
In this paper, we present our participation in CLEF Consumer Health Search Task 2018, mainly, its monolingual and multilingual subtasks: IRTask1 and IRTask4. In IRTask1, we use language-model based retrieval model, vector-space model and Kullback-Leiber divergence query expansion mechanism to build our runs. In IRTask4, we submitted 4 runs for each language of Czech, French and German. We follow query-translation approach in which we employ a Statistical Machine Translation (SMT) system to get a ranked list of translation hypotheses in English. We use this list for two systems: the first one uses 1-best-list translation to construct queries, and the second one uses a hypotheses reranker to select the best translation (in terms of retrieval performance) to construct queries. We also present our term reranking model for query expansion, in which we deploy feature set from different resources (the document collection, Wikipedia articles, translation hypotheses). These features are used to train a logistic regression model that can predict the performance when a candidate term is added to a base query.
Tato zpráva popisuje účast týmu Univerzity Karlovy v Praze na CLEF eHealth Consumer Health Search Task 2018
[ "Shadi Saleh", "Pavel Pecina" ]
https://www.semanticscholar.org/paper/e3e20b48acbbc3206d09ebc90b4fef6dc02341c5/
null
null
null
null
null
null
null
null
null
null
null
null
cs
2018
Hegel's Holiday Boogie
Hegelovo prázdninové boogie
Hegel's Holiday Boogie or Mondrian, Magritte and Avantgarde Dialectics In this talk we expore hegelian inspiration in Mondrian's theory of art and Magritte's paintings and propose to interpret the development of avantgarde painting as dialectics of the real and the reasonable.
Hegelovo prázdninové boogie aneb Mondrian, Magritte a avantgardní dialektika Avantgarda, jako umění, které často tematizuje svoji vlastní formu, je z hlediska Hegelovy filosofie zajímavá jako příklad rozvoje sebe-vědomí. V tomto přípěvku se však naopak podíváme na to, proč avantgardu zajímala Hegelova filosofie. Na příkladech Mondrianovy abstrakce a Magrittova surrealismu ukážeme dva diametrálně odlišné přístupy k inspiraci Hegelem ve výtvarném umění. Piet Mondrian se na Hegela odkazuje ve svých textech o teorii malířství. Pod vlivem hegelovské filosofie dospívá ke koncepci zobrazování, které označuje jako „abstraktně-reálné“, výraz „oživené skutečnosti abstraktního“. René Magritte se k Hegelově filosofii vyjadřuje přímo názvy a obsahem svých obrazů. V některých z nich explicitně odkazuje k dialektice, která je však implicitně přítomná ve většině jeho díla. Na závěr se zamyslíme nad tím, jak můžeme vývoj avantgardního malířství interpretovat jako dialektiku rozumného a skutečného.
[ "Tomáš Musil" ]
https://www.semanticscholar.org/paper/3015d642592f33e0c7786e27784a46230f3b3187/
null
null
null
null
null
null
null
null
null
null
null
null
cs
2018
Neural Natworks and Computational Linguistics
Neuronové sítě a matematická lingvistika
Will computers ever be able to understand written text? Will they be able to translate it from English to Czech without mistakes? Will they be able to describe a picture? What are artificial neural networks and how are they connected to these questions?
Budou někdy počítače schopny pochopit psaný text? Přeložit ho bezchybně třeba z angličtiny do češtiny? Popsat co je na obrázku? Co jsou umělé neuronové sítě a jak mohou pomoci vyřešit tyto otázky? To se dozvíte v této přednášce. Když bude čas, zmíníme i jak neuronky hrají go nebo řídí auta.
[ "Tomáš Musil" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
cs
2018
Deep neural networks: when computers learn the human language
Hluboké neuronové sítě: když se počítače učí lidský jazyk
Presentation of word embeddings and related research at ÚFALu within Jeden den s informatikou a matematikou.
Prezentace word embeddings a souvisejícího výzkumu probíhajícího na ÚFALu v rámci Jednoho dne s informatikou a matematikou.
[ "Tomáš Musil", "Rudolf Rosa", "Jindřich Libovický" ]
https://www.semanticscholar.org/paper/fc27df9eed52f7378b67f8cd0648ef02ffdce179/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
A Vision for (Computational) Linguistics
Vize pro (počítačovou) lingvistiku
A brief opening presentation for a panel on AI and the Humanities
Stručný úvod k panelu o umělé inteligenci a humanitních oborech představuje možnosti neuronového strojového překladu a hlubokého učení obecně pro lingvistiku.
[ "Ondřej Bojar" ]
https://www.semanticscholar.org/paper/2a6578411c6bf49baf0df6f4de597aa2925c7afa/
null
null
null
null
null
null
null
null
null
null
null
null
cs
2018
When neural MT will replace translators?
Kdy neuronový strojový překlad nahradı́ překladatele?
Machine translation is the one of the key application of artificial inteligence and computational linguistics. Two revolutions happened recently: in the 1990's and these days. What's the current state of the art? Finally a chance that machines will understand the text?
Strojový překlad je královskou disciplínou umělé inteligence a počítačové lingvistiky. Jednu revoluci zažil v 90. letech minulého století (zavedení statistických metod), druhou zažívá právě teď, když neuronové sítě skokově zlepšily kvalitu překladu. Jak se nyní překlad dělá? Podívejme se na ten zázrak technicky, bez silných marketingových slov. Je snad konečně naděje, že počítače začnou textu rozumět?
[ "Ondřej Bojar" ]
https://www.semanticscholar.org/paper/7c4cb950ab8aa998efa5ba60c417b8cbebb5dc86/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
Neural Machine Translation: From Basics to Semiotics
Neuronový strojový překlad: Od základů k sémiotice
In my talk, I will highlight the benefit that neural machine translation (NMT) has over previous statistical approaches to MT. I will then present the current state of the art in neural machine translation, briefly describing the current best architectures and their performance and limitations. In the second part of the talk, I will outline my planned search for correspondence between sentence meaning as traditionally studied by linguistics (or even semantics and semiotics) and the continuous representations learned by neural networks.
Ve svém příspěvku podrobně vysvětlím, jakou hlavní výhodu má neuronový překlad oproti předešlým statistickým přístupům k MT.
[ "Ondřej Bojar" ]
https://www.semanticscholar.org/paper/03ec7f75a2c7cb644ec88949e98ba55316fe79c3/
null
null
null
null
null
null
null
null
null
null
null
null
cs
2018
Computer linguistics and machine translation: A whole new world through deep learning
Počítačová lingvistika a strojový překlad: Zcela nový svět díky hlubokému učení
An invited talk for Master students of FJFI at a seminar introducing various interesting research topics.
Zvaná přednáška na seminář pořádaný na FJFI, představující zajímavá vědecká témata studentům magisterského programu. Ve své přednášce jsem představil počítačovou lingvistiku a zejména neuronový strojový překlad.
[ "Ondřej Bojar" ]
https://www.semanticscholar.org/paper/3d0fd9bf646f280b447e6dd9ffb29993133b2a94/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
Promises and Pitfalls of Neural MT
Sliby a úskalí Neural MT
In my talk, I presented the current translation quality achieved by neural machine translation and thoroughly discussed the expected benefits and risks for interpreters arising from the EU project ELITR.
V prezentaci jsem představil současnou úroveň kvality dosahovanou neuronovým strojovým překladem a podrobně se v diskusi věnoval očekávaným výhodám a rizikům, která do překladatelské profese přináší projekt ELITR.
[ "Ondřej Bojar" ]
https://www.semanticscholar.org/paper/1ae61445ddaeb101c5f398e6951fb6b41dd82aa0/
null
null
null
null
null
null
null
null
null
null
null
null
cs
2018
Does Neural Machine Translation Understand the Translated Text?
Rozumí neuronový strojový překlad tomu, co překládá?
Deep machine learning has crucially changed the techniques used in machine translation in 2016. In my short talk, I will summarize the quick development and questions for the field for the next week and the next ten years.
Hluboké strojové učení v roce 2016 zásadním způsobem změnilo techniky používané ve strojovém překladu. Kvalita strojových překladů se skokově zlepšila a strmý růst stále ještě pozorujeme. Neuronové sítě slibují možnost "učit se reprezentacím". Daří se je tedy učit reprezentacím, které by ladily s něčím jako je význam vět? V krátkém příspěvku představím překotný vývoj na poli strojového překladu a otázky pro tento obor na příští týden i příštích deset let.
[ "Ondřej Bojar" ]
https://www.semanticscholar.org/paper/00dc987d9a4c5948c165330153888998e8a13ca9/
null
null
null
null
null
null
null
null
null
null
null
null
cs
2018
Artificial Intelligence Replacing Translators
Umělá inteligence nahrazuje překladatele
A quick introduction to prospective students of Computer Science and Computational Linguistics with recent advances in machine translation
Seznámení zájemců o obor informatiky a počítačové lingvistiky s nejnovějším vývojem na poli strojového překladu.
[ "Ondřej Bojar" ]
https://www.semanticscholar.org/paper/96e2bc6d7806d57eb01c5c743c46c0d6e46e1157/
null
null
null
null
null
null
null
null
null
null
null
null
cs
2018
Batch translation for IBM
Dávkový překlad IBM
Automated system for batch translation for IBM. The first supported translation direction was from English into Czech, other languages (e.g. Hungarian, Arabic) are added as needed.
Automatizovaný systém pro dávkový překlad pro IBM. Prvním jazykovým párem byl překlad z angličtiny do češtiny, další jazyky (maďarština, arabština ap.) jsou přidávány dle potřeby.
[ "Ondřej Bojar", "Jan Hajič" ]
https://www.semanticscholar.org/paper/e55607c24c76a81d12421428e378f1d0d8e6242d/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
EvalD Reference-Less Discourse Evaluation for WMT18
EvalD, diskurzní evaluace bez referenčního překladu pro WMT18
We present results of automatic evaluation of discourse in machine translation (MT) outputs using the EVALD tool, showing that automatic evaluation of discourse in translated texts allows for distinguishing individual MT systems.
Užíváme systém EVALD k automatickému ohodnocení výstupů několika systémů strojového překladu; výsledky ukazují, že automatické ohodnocení diskurzu v textech umožňuje rozlišit různé systémy strojového překladu.
[ "Ondřej Bojar", "Jiří Mírovský", "Kateřina Rysová", "Magdaléna Rysová" ]
https://www.semanticscholar.org/paper/6f423281b52e5511bcdc0bab70e1fab4796d4f91/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
Findings of the 2018 Conference on Machine Translation (WMT18)
Výsledky konference ve strojovém překladu (WMT18)
This paper presents the results of the premier shared task organized alongside the Confer- ence on Machine Translation (WMT) 2018. Participants were asked to build machine translation systems for any of 7 language pairs in both directions, to be evaluated on a test set of news stories. The main metric for this task is human judgment of translation quality. This year, we also opened up the task to additional test suites to probe specific aspects of transla- tion.
Článek přináší výsledky hlavní společné úlohy organizované při konferenci o strojovém překladu (WMT) v roce 2018.
[ "Ondřej Bojar", "Christian Federmann", "Mark Fishel", "Yvette Graham", "Barry Haddow", "Matthias Huck", "Philipp Koehn", "Christof Monz" ]
https://www.semanticscholar.org/paper/d098bdca6d0335a31fd164c1fac68028784a0737/
null
null
null
null
null
null
null
null
null
null
null
null
cs
2018
Prague Dependency Treebank 3.5
Prague Dependency Treebank 3.5
The Prague Dependency Treebank 3.5 is the 2018 edition of the core Prague Dependency Treebank (PDT). It contains all PDT annotation made at the Institute of Formal and Applied Linguistics under various projects between 1996 and 2018 on the original texts, i.e., all annotation from PDT 1.0, PDT 2.0, PDT 2.5, PDT 3.0, PDiT 1.0 and PDiT 2.0, plus corrections, new structure of basic documentation and new list of authors covering all previous editions. The Prague Dependency Treebank 3.5 (PDT 3.5) contains the same texts as the previous versions since 2.0; there are 49,431 annotated sentences (over 800 thousand nodes) on all layers, from tectogrammatical to words, and additional sentences on the analytical (surface dependency syntax) and morphological layers of annotation (approx. 2 million words in total). Closely linked to the tectogarammtical layer is the annotation of sentence information structure, multiword expressions, coreference, bridging relations and discourse relations.
Závislostní korpus anotovaný na rovině morfologické (2 miliony slov), syntaktické (1,5 milionu slov) a sémantické (přes 0,8 milionu slov, tedy 49,5 tis. vět). Obsahuje všechny PDT anotace originálních textů, které vznikly v rámci různých projektů na UFALu mezi roky 1996 a 2018 (PDT 1.0, PDT 2.0, PDT 2.5, PDT 3.0, PDiT 1.0 and PDiT 2.0) a jejich opravy; seznam autorů pokrývá autory všech publikovaných korpusů.
[ "Jan Hajič", "Eduard Bejček", "Alevtina Bémová", "Eva Buráňová", "Eva Hajičová", "Jiří Havelka", "Petr Homola", "Jiří Kárník", "Václava Kettnerová", "Natalia Klyueva", "Veronika Kolářová", "Lucie Kučová", "Markéta Lopatková", "Marie Mikulová", "Jiří Mírovský", "Anna Nedoluzhko", "Petr Pajas", "Jarmila Panevová", "Lucie Poláková", "Magdaléna Rysová", "Petr Sgall", "Johanka Spoustová", "Pavel Straňák", "Pavlína Synková", "Magda Ševčíková", "Jan Štěpánek", "Zdeňka Urešová", "Barbora Vidová Hladká", "Daniel Zeman", "Šárka Zikánová", "Zdeněk Žabokrtský" ]
https://www.semanticscholar.org/paper/c2e9552a6d3f683469e547cfba0841ff465c0c50/
null
null
null
null
null
null
null
null
null
null
null
null
cs
2018
Diathesis and recoprocity (on Czech material)
Diateze a reciprocita (na materiálu češtiny)
The short survey of the constructions understood as grammatical diathesis is given. The discussion of the generally accepted hypothesis that the surface expression of the actor is excluded in deagentive constructions. The counterexamples where the combination of the deagentive and reciprocal diathesis are submitted. Some restrictions for such constructions are studied as well.
Ve stati se předládají vedle přehledu konstrukcí pokládaných za gramatickou diatezi argumenty pro případy, kdy deagentní diateze spojená s reciproční diatezí může vyjádřit konatele pomocí předložkového výrazu mezi + Instrumentál.Jsou zde uvedena jistá omezení na tvorbu těchto konstrukcí. Ilustruje se také jejich častá homonymie s konstrukcemi s významem lokace.
[ "Jarmila Panevová" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
cs
2018
A Competition of the Prepositional Cases in the Space Constructions in Czech
Konkurence předložkových pádů u jádra lokálních určení v češtině
The distribution of the Czech prepositional constructions connected with the semantics of space is studied and the proposal how to present their semantics within the grammatical description is submitted. The prepositional constructions v+6, na+6, u+2, and po+6 are included, the rich data for this analysis were obtained from PDT 3.0 using the tool ForFun 1.0
V článku se podrobují diskusi distinkce uvnitř lokálního určení spojené s distribucí výrazů v předložkových konstrukcích v + Loc, na + Loc, u + Gen, po + 6 a s jejich sémantickými specifiky. Stať je stavěna na materiálu získaného pomocí nástroje ForFun 1.0.
[ "Jarmila Panevová" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
cs
2018
An Actant or a Free Modification? (On Untypical Forms in Valency Patterns of Nouns)
Aktant, nebo volné doplnění? (K netypickým formám ve valenčním poli substantiv)
In this paper, we demonstrate how to treat some difficult phenomena related to non-standard expression of valency patterns of Czech deverbal nouns. We introduce some new attributes to indicate (i) alternative, stylistically marked forms of actants expressed by secondary prepositions, and (ii) marginal forms of actants expressed by primary prepositions. We show that the number of forms of adnominal actants is often increased in comparison with the verbal ones, differentiating regular forms and alternative forms. We also discuss cases of ambiguous forms of two different actants and give criteria for their interpretation.
Příspěvek je věnován několika problémům, které vznikají při aplikaci valenční teorie FGP na valenci deverbativních substantiv a které demonstrují fakt, že aplikace valence slovesné na substantivní nemůže být přímočará a že substantivní valenční rámce nejsou vždy analogií rámců jejich základových sloves. U substantiva přibývají formy typické pro volná doplnění uvozené zpravidla sekundárními předložkami, které se svou lexikální náplní obsahově blíží obligatorním aktantům (oznámení ze strany úřadu ohledně návštěvních hodin). Pokládáme je za stylové varianty, které se nezapisují do valenčních rámců, ale jsou ošetřeny speciálními pravidly. Jiné, uvozené primárními předložkami, mají marginální povahu; vyžadují však zjištění kontextových podmínek pro jejich užití (otázka nad stavem společnosti / po stavu společnosti), včetně podmínek jejich vzájemné zaměnitelnosti, popř. též úvah nad odstraněním jejich homonymie (otázka na psychologa.ADDR/PAT).
[ "Jarmila Panevová", "Veronika Kolářová" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
What We Can Learn From J.-M. Zemb About Negation and Information Structure: A View From Prague
Co se můžeme naučit z díla J.- M. Zembyho o negaci a aktuálním členění: Pohled z Prahy
A critical analysis of the writings of a prominent French linguist J.-M. Zemb, especially those concerning the information structure and negation, from the point of view of the author's own approach to these basic issues.of the syntactic structure of the sentence.
Kritický rozbor díla předního francouzského lingvisty J.-M. Zemba, především jeho prací o informační struktuře věty a negaci z pohledu autorčina vlastního přístupu k těmto základním otázkám syntaktické struktury. věty.
[ "Eva Hajičová" ]
https://www.semanticscholar.org/paper/ea30138b3e0a30f2d40eb592f3f9fa2985b5df80/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
Discourse Coherence Through the Lens of an Annotated Text Corpus: A Case Study
Koherence textu pohledem anotovaného korpusu: případová studie
A corpus-based study of local coherence as established by anaphoric links between the elements in the thematic (Topic) and the rhematic (Focus) parts of sentences in different genres of texts in the Prague Dependency Treebank.
Korpusová studie lokální koherence textu založené na anaforických vztazích mezi elementy v tématické (Topic) a rématické (Focus) části věty v různých žánrech textů Pražského závislostního korpusu.
[ "Eva Hajičová", "Jiří Mírovský" ]
https://www.semanticscholar.org/paper/0ddc59374aa6133aefd5378665126188a336a7a2/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
Identification of Thematic Discourse Relations on the Data from an Annotated Corpus of Czech
Identifikace tématických diskurzních vztahů v datech anotovaného korpusu češtiny
We analyze the data of the Prague Discourse Treebank 2.0 as for the text coherence based on the so-called thematic progressions, that is chains of anaphoric links between sentences with regard to their topic–focus articulation (information structure).
Analyzujeme data Pražského diskurzního korpusu 2.0 a na jejich základě studujeme tzv. tématické posloupnosti, tj. řetězce anaforických vztahů mezi větami vzhledem k jejich informační struktuře (topic–focus articulation).
[ "Eva Hajičová", "Jiří Mírovský" ]
https://www.semanticscholar.org/paper/434c239171782533f4fb069b80571df55f0b71cd/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
Topic/Focus vs. Given/New: Information Structure and Coreference Relations in an Annotated Corpus
Topic/Focus vs. Given/New: Informační struktura věty a koreferenční vztahy v anotovaném korpusu
We put under scrutiny two dichotomies discussed in the information structure literature, namely the dichotomy of topic and focus based on the relation of aboutness, and the dichotomy between given and new information. In particular, we examine whether the topic/focus dichotomy can be based on the distinction between given and new information, or whether the ‘aboutness’ relation is a more appropriate basis.
Zkoumáme dvě dichotomie, kterými se zabývá literatura o informační struktuře věty, konkrétně dichotomii topic/focus založenou na vztahu "býti o čem", a dichotomii mezi danou a novou informací. Zaměřujeme se obzvláště na otázku, zda dichotomie topic/focus může být založena na rozlišení mezi danou a novou informací, či zda vztah "býti o čem" tvoří přesnější obraz situace.
[ "Eva Hajičová", "Jiří Mírovský" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
A Cross-lingual Synonym Classes Lexicon
Bilingvní slovník synonymních tříd
In the present paper, we focus on synonymy of verbs in a bilingual, Czech-English setting. We introduce a new lexical resource called CzEngClass. Our research of semantic equivalence of verbs is based on the FGD theory on the syntactic side, and gets main inspiration from FrameNet on the semantic side. As the main evidence, we use a parallel dependency corpus (the Prague Czech-English Dependency Treebank 2.0). We consider this “bottom-up” approach a novel and appropriate approach to study verbal synonymy. Synonymous Czech and English verbs are being grouped into cross-lingual synonym classes and captured in the CzEngClass lexicon. This lexicon contains not only mappings of valency arguments to semantic roles for each member of the synonym group, but also links them to individual verb entries as captured in selected existing lexical resources, such as FrameNet, VerbNet, Vallex(es) and Czech and English WordNets. We believe such explicit description of verbal synonymy relations, collected in a richly interconnected lexicon, contributes valuable knowledge to both traditional linguistics and various NLP tasks.
Příspěvek o slovníku anglicko-českých slovesných synonym na základě PCEDT. Synonyma jsou řazena do synonymních tříd, obsahujících anglické a české členy, které jsou zároveň propojeny s externími a interními lexikálními zdroji.
[ "Zdeňka Urešová", "Eva Fučíková", "Eva Hajičová" ]
https://www.semanticscholar.org/paper/585180be10e9fc6c482b2f7f50b77b9976acfa0c/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
CzEngClass 0.1
CzEngClass 0.1
The CzEngClass synonym verb lexicon is a result of a project investigating semantic ‘equivalence’ of verb senses and their valency behavior in parallel Czech-English language resources, i.e., relating verb meanings with respect to contextually-based verb synonymy. The lexicon entries are linked to PDT-Vallex (http://hdl.handle.net/11858/00-097C-0000-0023-4338-F), EngVallex (http://hdl.handle.net/11858/00-097C-0000-0023-4337-2), CzEngVallex (http://hdl.handle.net/11234/1-1512), FrameNet (https://framenet.icsi.berkeley.edu/fndrupal/), VerbNet (http://verbs.colorado.edu/verbnet/index.html), PropBank (http://verbs.colorado.edu/%7Empalmer/projects/ace.html), Ontonotes (http://verbs.colorado.edu/html_groupings/), and Czech (http://hdl.handle.net/11858/00-097C-0000-0001-4880-3) and English Wordnets (https://wordnet.princeton.edu/). Part of the dataset is a file reflecting annotators choices for assignment of verbs to classes.
Synonymní slovník CzEngClass je výsledkem projektu zkoumajícího sémantickou „ekvivalenci“ slovesných významů a jejich valenčního chování v paralelních zdrojích česko-anglického jazyka, tj. vztahujících se k slovním významům s ohledem na kontextuálně založená slovesná synonyma. Položky lexikonu jsou propojeny s PDT-Vallexem (http://hdl.handle.net/11858/00-097C-0000-0023-4338-F), EngVallexem (http://hdl.handle.net/11858/00-097C-0000-0023-4337-2), CzEngVallexem (http://hdl.handle.html.net/11234/1-1512), FrameNetem (https://framenet.icsi.berkeley.edu/verbverbands/), VerbNetem (http://coloredu.html.html Součástí datového souboru je soubor odrážející volby anotátorů pro přiřazení sloves do tříd.
[ "Zdeňka Urešová", "Eva Fučíková", "Eva Hajičová", "Jan Hajič" ]
https://www.semanticscholar.org/paper/97c897bf3e035a728600cd412751e2a2bddd2392/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
CzEngClass 0.2
CzEngClass 0.2
The CzEngClass synonym verb lexicon is a result of a project investigating semantic ‘equivalence’ of verb senses and their valency behavior in parallel Czech-English language resources, i.e., relating verb meanings with respect to contextually-based verb synonymy. The lexicon entries are linked to PDT-Vallex (http://hdl.handle.net/11858/00-097C-0000-0023-4338-F), EngVallex (http://hdl.handle.net/11858/00-097C-0000-0023-4337-2), CzEngVallex (http://hdl.handle.net/11234/1-1512), FrameNet (https://framenet.icsi.berkeley.edu/fndrupal/), VerbNet (http://verbs.colorado.edu/verbnet/index.html), PropBank (http://verbs.colorado.edu/%7Empalmer/projects/ace.html), Ontonotes (http://verbs.colorado.edu/html_groupings/), and Czech (http://hdl.handle.net/11858/00-097C-0000-0001-4880-3) and English Wordnets (https://wordnet.princeton.edu/). Part of the dataset are files reflecting annotators choices and agreement for assignment of verbs to classes.
Synonymní slovník CzEngClass je výsledkem projektu zkoumajícího sémantickou „ekvivalenci“ slovesných významů a jejich valenčního chování v paralelních zdrojích česko-anglického jazyka, tj. vztahujících se k slovním významům s ohledem na kontextuálně založená slovesná synonyma. Položky lexikonu jsou propojeny s PDT-Vallexem (http://hdl.handle.net/11858/00-097C-0000-0023-4338-F), EngVallexem (http://hdl.handle.net/11858/00-097C-0000-0023-4337-2), CzEngVallexem (http://hdl.handle.html.net/11234/1-1512), FrameNetem (https://framenet.icsi.berkeley.edu/verbverbands/), VerbNetem (http://coloredu.html.html Součástí datového souboru je soubor odrážející volby anotátorů pro přiřazení sloves do tříd.
[ "Zdeňka Urešová", "Eva Fučíková", "Eva Hajičová", "Jan Hajič" ]
https://www.semanticscholar.org/paper/f02f4b5eb5ffe053df68b79bb0ad829092d0b989/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
A CROSS-LINGUAL SYNONYM CLASSES LEXICON
Slovník vícejazyčných synonymních tříd
In the present article, we focus on valency and synonymy of verbs in a bilingual, Czech-English setting. Our research of semantic equivalence of verbs is based on the FGD theory on the syntactic side (including valency), and gets main inspiration from FrameNet and VerbNet on the semantic side. As the main source of evidence, we use the Prague Czech-English Dependency Treebank 2.0. We consider this “bottom-up” approach a novel and appropriate approach to study verbal synonymy. Synonymous Czech and English verbs are being grouped into cross-lingual synonym classes and captured in the new CzEngClass lexicon. This lexicon contains not only mappings of valency arguments to semantic roles for each member of the synonym group, but also links them to individual verb entries in FrameNet, VerbNet, Vallex(es) and Czech and English WordNets, making CzEngClass also a richly interconnected lexicon.
V článku se zaměřujeme na valency a synonymii sloves v bilingvním česko-anglickém kontextu. Náš výzkum sémantické ekvivalence sloves je z pohledu syntaktického (včetně valence) založen na FGP a z pohledu sémantického je inspirován především FrameNetem a VerbNetem. Jako hlavní zdroj korpusových dokladů používáme Pražský česko-anglický závislostní korpus 2.0. Postup výzkumu “od spoda nahoru” považujeme za nový a adekvátní přístup pro stadium slovesné synonymie. Synonymní česká a anglická slovesa jsou uskupeny do mezijazykových synonymních tříd v novém slovníku CzEngClass. Tento slovník pro každé synonymní sloveso dané třídy obsahuje jak mapování valenčních argumentů na sémantické role, tak propojení s jednotlivými slovesnými významy ve FrameNetu, VerbNetu, ve Vallexech a v anglickém WordNetu, což ze slovníku CzEngClass zároveň vytváří bohatě propojený slovník.
[ "Zdeňka Urešová", "Eva Fučíková", "Eva Hajičová", "Jan Hajič" ]
https://www.semanticscholar.org/paper/585180be10e9fc6c482b2f7f50b77b9976acfa0c/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
Creating a Verb Synonym Lexicon Based on a Parallel Corpus
Slovesný synonymický slovník na základě paralelního korpusu
This paper presents the first findings of our recently started project of building a new lexical resource called CzEngClass, which consists of bilingual verbal synonym groups. In order to create such a resource, we explore semantic ‘equivalence’ of verb senses of generally different verbs in a bilingual (Czech-English) setting by using translational context of real-world texts in a parallel, richly annotated dependency corpus. When grouping semantically equivalent verb senses into classes of synonyms, we focus on valency (arguments as deep dependents with morphosyntactic features relevant for surface dependencies) and its mapping to a set of semantic “roles” for verb arguments, common within one class. We argue that the existence of core argument mappings and certain adjunct mappings to a common set of semantic roles is a suitable criterion for a reasonable verb synonymy definition, possibly accompanied with additional contextual restrictions. By mid-2018, the first version of the lexicon called CzEngClass will be publicly available.
Představujeme první výsledky našeho projektu o slovesné synonymii na základě paralelního česko-anglického korpusu.
[ "Zdeňka Urešová", "Eva Fučíková", "Eva Hajičová", "Jan Hajič" ]
https://www.semanticscholar.org/paper/a81b633ac8e3caf1727a69344230fb7713b2cea4/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
Defining Verbal Synonyms: between Syntax and Semantics
Definování slovesných synonym: mezi syntaxí a sémantikou
While studying verbal synonymy, we have investigated the relation between syntax and semantics in hope that the exploration of this relationship will help us to get more insight into the question of synonymy as the relationship relating (similar) meanings between different lexemes. Most synonym lexicons (Wordnets and similar thesauri) are based on an intuition about the similarity of word meanings, or on notions like “semantic roles.” In some cases, syntax is also taken into account, but we have found no annotation and/or evaluation experiment to see how strongly can syntax contribute to synonym specification. We have prepared an annotation experiment for which we have used two treebanks (Czech and English) from the Universal Dependencies (UD) set of parallel corpora (PUDs) in order to see how strong correlation exists between syntax and the assignment of verbs in context to pre-determined (bilingual) classes of synonyms. The resulting statistics confirmed that while syntax does support decisions about synonymy, such support is not strong enough and that more semantic criteria are indeed necessary. The results of the annotation will also help to further improve rules and specifications for creating synonymous classes. Moreover, we have collected evidence that the annotation setup that we have used can identify synonym classes to be merged, and the resulting data (which we plan to publish openly) can possibly serve for the evaluation of automatic methods used in this area.
Při studiu synonymických vztahů jsme vzali v potaz vztah syntaxe a sémantiky, abychom mohli lépe definovat slovesnou valenci. Za pomocí anotačního experimentu na datech UD a PUDs jsme zjišťovali koreleaci syntaxe a sémantiky.
[ "Zdeňka Urešová", "Eva Fučíková", "Eva Hajičová", "Jan Hajič" ]
https://www.semanticscholar.org/paper/cf60fb09522c751688d8c60e2c544785d540c80a/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
Synonymy in Bilingual Context: The CzEngClass Lexicon
Synonymie v bilingvním kontextu: Slovník CzEngClass
This paper describes CzEngClass, a bilingual lexical resource being built to investigate verbal synonymy in bilingual context and to relate semantic roles common to one synonym class to verb arguments (verb valency). In addition, the resource is linked to existing resources with the same or a similar aim: English and Czech WordNet, FrameNet, PropBank, VerbNet (SemLink), and valency lexicons for Czech and English (PDT-Vallex, Vallex and EngVallex). There are several goals of this work and resource: (a) to provide gold standard data for automatic experiments in the future (such as automatic discovery of synonym classes, word sense disambiguation, assignment of classes to occurrences of verbs in text, coreferential linking of verb and event arguments in text, etc.), (b) to build a core (bilingual) lexicon linked to existing resources, for comparative studies and possibly for training automatic tools, and (c) to enrich the annotation of a parallel treebank, the Prague Czech English Dependency Treebank, which so far contained valency annotation but has not linked synonymous senses of verbs together. The method used for extracting the synonym classes is a semi-automatic process with a substantial amount of manual work during filtering, role assignment to classes and individual members’ arguments, and linking to the external lexical resources. We present the first version with 200 classes (about 1800 verbs) and evaluate interannotator agreement using several metrics.
V článku popisujeme slovník bilingvní slovník anglických a českých slovesných synonym, CzEngClass. Ta jsou zachycena v synonymních třídách na základě kontextu. Zároveň je jejich valenční struktura vztažena k sémantickým rolím přiřazeným dané synonymní třídě. Navíc jsou jednotlivé členy synonymní třídy propojeny s externími (FrameNet, VerbNet, PropbBank, WordNet) a interními zdroji (PDT-valency lexicons, Vallex and Czech WordNet). V čánku představujeme první verzi slovníku, obsahující 200 tříd (cca 1800 sloves) a evaluaci mezianotátorské shody.
[ "Zdeňka Urešová", "Eva Fučíková", "Eva Hajičová", "Jan Hajič" ]
https://www.semanticscholar.org/paper/d77056078a7af8650b90df2eca90cd7e30098752/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
Tools for Building an Interlinked Multilingual Synonym Lexicon Network
Nástroje pro vytvoření synonymního slovníku
This paper (poster) presents a tool used for building a new cross-lingual verbal synonym lexical resource called CzEngClass, as well as the structure of the lexicon. This lexicon captures interlingual synonyms, using valency behavior of synonymous verbs in relation to semantic roles as one of the criteria for defining such interlingual synonymy. The tool, called Synonym Class Editor - SynEd, is a user-friendly tool specifically customized to build and edit individual entries in the lexicon. It helps to keep the cross-lingual synonym classes consistent and linked to internal as well as well-known external lexical resources. The structure of SynEd also allows to keep and edit the appropriate syntactic and semantic information for each Synonym Class member. The editor makes it also possible to display examples of class members usage in translational (parallel corpus) context. SynEd runs platform independently and may be used for multiple languages. SynEd, CzEngClass and services based on them will be openly available.
Tento příspěvek představuje editor pro tvorbu bilingvního slovesného synonymického slovníku, který propojuje jednotlivé slovesné třídy a jejich členy s dalšími lexikálními zdroji.
[ "Zdeňka Urešová", "Eva Fučíková", "Eva Hajičová", "Jan Hajič" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
cs
2018
Adnominal Valency Complementations Expressed by Prepositional Groups
Předložková vyjádření adnominálních valenčních doplnění
Prepositional groups represent the second most frequent form of valency complementations of Czech nouns (preceded only by prepositionless genitive). In this paper, we exploit the NomVallex lexicon which covers nouns belonging to three semantic classes, i.e. Communication, Mental action and Psychological state. We show that adnominal prepositional groups are shared among these semantic classes to a large extent but their frequency differs. A comparison between valency of nouns from NomVallex and valency of their base verbs from Vallex lexicon reveals that various non-systemic shifts in surface forms of participants exist between the verbal and adnominal prepositional groups. It also enables us to state a high ratio of non-systemic adnominal prepositional groups to all adnominal prepositional groups in the lexicon (not less than 39%).
Předložkové skupiny představují druhou nejčastější formu valenčních doplnění českých substantiv. V této studii byla zkoumána na datech valenčního slovníku NomVallex, a to u tří sémantických tříd, Communication, Mental action a Psychological state. Tyto tři třídy adnominální předložkové skupiny ve velké míře sdílejí, liší se však míra jejich zastoupení. Srovnání valenčních vlastností substantiv z NomVallexu a jejich základových sloves z valenčního slovníku Vallex nám umožnilo upřesnit popis protějšků adnominálních i adverbálních předložkových skupin a ukázat, že tyto předložkové skupiny podléhají nejrůznějším posunům ve formách participantů. Díky tomuto srovnání byl také zjištěn vysoký podíl nově získaných adnominálních předložkových skupin v celkovém počtu předložkových skupin (minimálně 39%).
[ "Veronika Kolářová", "Anna Vernerová", "Jana Klímová" ]
https://www.semanticscholar.org/paper/55bc24142fafdf79de253ae6ea919df54b66b805/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
Valency Lexicon of Czech Nouns NomVallex: Starting Point and Goals
Valenční slovník českých substantiv NomVallex: Východiska a cíle
The corpus-based valency lexicon of Czech nouns NomVallex is a starting project building upon the theory of valency developed within Functional Generative Description and extending the two lexicons developed within this tradition, PDT-Vallex and Vallex. The linguistic material is extracted from Czech linear and syntactically annotated corpora. In comparison with PDT-Vallex, the treatment of entries will be more exhaustive, for example, in the coverage of senses and in the semantic classification added to selected lexical units. The main criteria for including nouns in the lexicon will be semantic class membership and the complexity of valency patterns. Valency of nouns will be captured in the form of valency frames, enumeration of combinations of adnominal participants, and corpus examples.
České valenční slovníky doposud pokrývaly především valenci slovesnou, slovníkové zpracování valence substantiv se tedy potýká s řadou teoretických i praktických otázek. V příspěvku představujeme východiska a cíle při budování valenčního slovníku českých deverbativních substantiv zvaného NomVallex, který vzniká v rámci projektu Valenční slovník substantiv založený na korpusu, s počátkem řešení v lednu 2016. Po teoretické stránce vycházíme z valenční teorie funkčního generativního popisu, při vlastním lexikografickém zpracování se opíráme o korpusový materiál (data Pražského závislostního korpusu a Českého národního korpusu – subkorpusů řady SYN). Při tvorbě slovníku budou částečně využita data valenčního slovníku PDT-Vallex, zpracování hesel však bude zdokonaleno v několika aspektech (například v pokrytí všech významů substantiva). Kritériem pro zařazení do slovníku bude především sémantická třída substantiva a složitost jeho valenčního chování, zejména specifické formy participantů. Valence substantiv bude zachycena pomocí valenčních rámců, výčtu všech možných kombinací adnominálních participantů a korpusových příkladů.
[ "Veronika Kolářová", "Jana Klímová", "Anna Vernerová" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
EVALD 3.0 for Foreigners – Evaluator of Discourse
EVALD 3.0 pro cizince – Evaluátor diskurzu
Evaluator of Discourse for Foreigners (EVALD 3.0 for Foreigners) is a software that serves for automatic evaluation of surface coherence (cohesion) in Czech texts written by non-native speakers of Czech. Software evaluates the level of the given text from the perspective of its surface coherence (i.e. it takes into consideration, e.g., the frequency and diversity of connectives, richness of vocabulary, coreference relations etc.). The new version (3.0) adds a set of features related to the topic-focus articulation.
Počítačový nástroj Evaluátor diskurzu pro cizince (EVALD 3.0 pro cizince) slouží k automatickému hodnocení povrchové koherence (koheze) textů v češtině psaných nerodilými mluvčími češtiny. Software hodnotí úroveň předloženého textu z hlediska jeho povrchové výstavby (zohledňuje např. frekvenci a rozmanitost užitých spojovacích prostředků, bohatost slovní zásoby, koreferenční vztahy apod.). Nová verze (3.0) přidává množinu rysů zaměřených na informační strukturu věty.
[ "Jiří Mírovský", "Michal Novák", "Kateřina Rysová", "Magdaléna Rysová", "Eva Hajičová" ]
https://www.semanticscholar.org/paper/8994bf0da31916b59fd67f36cda2afbd3ef74ac8/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
EVALD 3.0 – Evaluator of Discourse
EVALD 3.0 – Evaluátor diskurzu
Evaluator of Discourse (EVALD 3.0) is a software that serves for automatic evaluation of surface coherence (cohesion) in Czech texts written by native speakers of Czech. Software evaluates the level of the given text from the perspective of its surface coherence (i.e. it takes into consideration, e.g., the frequency and diversity of connectives, richness of vocabulary, coreference relations etc.). The new version (3.0) adds a set of features related to the topic-focus articulation.
Počítačový nástroj Evaluátor diskurzu (EVALD 3.0) slouží k automatickému hodnocení povrchové koherence (koheze) textů v češtině psaných rodilými mluvčími češtiny. Software tedy hodnotí (známkuje) úroveň předloženého textu z hlediska jeho povrchové výstavby (zohledňuje např. frekvenci a rozmanitost spojovacích prostředků, bohatost slovní zásoby, koreferenční vztahy apod.). Nová verze (3.0) přidává množinu rysů zaměřených na informační strukturu věty.
[ "Jiří Mírovský", "Michal Novák", "Kateřina Rysová", "Magdaléna Rysová", "Eva Hajičová" ]
https://www.semanticscholar.org/paper/531a6482158192e6ff302a86ce63b85bb83b3dfe/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
Core Arguments in Universal Dependencies
Základní aktanty v Universal Dependencies
We will look at one of the cornerstones of Universal Dependencies relation taxonomy, the core arguments and their difference from oblique dependents.
Podíváme se na jeden ze základních kamenů taxonomie vztahů v Universal Dependencies, na tzv. core arguments (základní aktanty) a na způsoby, jak je lze odlišit od tzv. oblique dependents (vedlejší aktanty a volná doplnění).
[ "Daniel Zeman" ]
https://www.semanticscholar.org/paper/6c3523cb0f0823c6754a1d0b8a25a4baf39d2edd/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
The World of Tokens, Tags and Trees
Svět tokenů, značek a stromů
This monograph presents a comparative study of annotation approaches to morphology and syntax of natural languages, with emphasis on applicability in a multilingual environment. Annotation is understood as adding linguistic categories and relations to digitally encoded natural language text, resulting in annotated corpus; as syntactic relations are often represented in the form of dependency trees, the annotated corpora covered by the monograph are dependency treebanks. Many treebanks exist and their annotation styles vary significantly, which hampers their usefulness for linguists and language engineers. We survey several harmonization efforts that tried to come up with cross-linguistically applicable annotation guidelines, including the most recent and broadest effort to date, Universal Dependencies. We examine language description on three levels: 1. tokenization and word segmentation, 2. morphology, and 3. surface dependency syntax. For each language phenomenon we provide a comparison of its analysis and annotation in various existing treebanks (or other corpora, for tokenization and morphology), pointing out advantages and disadvantages of the competing approaches. On the morphological layer, we go even beyond the currently available corpora and provide a typological survey of features that will be needed when less-resourced languages are covered by an annotation project. We conclude that no single approach is suitable for all purposes, but a good approach must not lose information, so that annotation can be converted to another style when necessary.
Tato monografie představuje srovnávací studii přístupů k anotaci morfologie a syntaxe přirozených jazyků s důrazem na použitelnost v mnohojazyčném prostředí. Anotací se rozumí přidání lingvistických kategorií a vztahů do digitalizovaného textu v přirozeném jazyce. Výsledkem je anotovaný korpus; vzhledem k tomu, že syntaktické vztahy jsou často reprezentované jako závislostní stromy, tato monografie se soustředí na závislostní korpusy (treebanky).
[ "Daniel Zeman" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
Uniformity versus Diversity: Dependency Grammar for All
Jednota vedle různosti: Závislostní gramatika pro všechny
I will present Universal Dependencies, a community project in computational linguistics that develops morphologically and syntactically annotated corpora for a large number of languages. UD started as an attempt to harmonize language resources used in various software tools for natural language processing. However, it quickly became a valuable resource also for corpus-based research. Three years after its first release, UD contains over 100 dependency treebanks in more than 60 languages. Despite the inevitable bias towards “big” languages, treebanks from less resourced language families are gradually added. In my talk, I will discuss some challenges of building a universal annotation scheme for all languages. In particular, how to make sure that comparable phenomena are annotated in a comparable fashion, without making all the languages look the same.
Představím Universal Dependencies, komunitní projekt v počítačové lingvistice, který poskytuje morfologicky a syntakticky anotované korpusy v mnoha jazycích.
[ "Daniel Zeman" ]
https://www.semanticscholar.org/paper/81ad07b3fee07c95deadab7c4a1cc146d761a0a9/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
Universal Dependencies and the Slavic Languages
Universal Dependencies a slovanské jazyky
I will present Universal Dependencies, a worldwide community effort aimed at providing multilingual corpora, annotated at the morphological and syntactic levels following unified annotation guidelines. I will discuss the concept of core arguments, one of the cornerstones of the UD framework. In the second part of the talk I will focus on some interesting problems and challenges of applying Universal Dependencies to the Slavic languages. I will discuss examples from 12 Slavic languages that are currently represented in UD and show that cross-linguistic consistency can still be improved.
Představím Universal Dependencies, celosvětový komunitní projekt zacílený na tvorbu mnohojazyčných korpusů anotovaných podle jednotných pravidel na morfologické a syntaktické rovině. Proberu koncept základních aktantů (core arguments), jednoho z pilířů, na kterých stojí anotační schéma UD. Ve druhé části přednášky se zaměřím na některé zajímavé problémy spojené s aplikací Universal Dependencies na slovanské jazyky. Na příkladech z 12 slovanských jazyků, které jsou v současné době reprezentované v UD, ukážu, že na mezijazykové konzistenci je ještě stále co zlepšovat.
[ "Daniel Zeman" ]
https://www.semanticscholar.org/paper/1379d2594b97bda4cc79c93c54a1d2d68867a744/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
CoNLL 2018 Shared Task System Outputs
Výstupy systémů, které se účastnily soutěže CoNLL 2018
Test data parsed by systems submitted to the CoNLL 2018 shared task in parsing Universal Dependencies.
Testovací data rozebraná systémy, které se účastnily soutěže v syntaktické analýze Universal Dependencies při konferenci CoNLL 2018.
[ "Daniel Zeman", "Martin Potthast", "Elie Duthoo", "Olivier Mesnard", "Piotr Rybak", "Alina Wróblewska", "Wanxiang Che", "Yijia Liu", "Yuxuan Wang", "Bo Zheng", "Ting Liu", "Zuchao Li", "Shexia He", "Zhuosheng Zhang", "Hai Zhao", "Yingting Wu", "Jia-Jun Tong", "Dat Quoc Nguyen", "Karin Verspoor", "Hui Wan", "Tahira Naseem", "Young-Suk Lee", "Vittorio Castelli", "Miguel Ballesteros", "Daniel Hershcovich", "Omri Abend", "Ari Rappoport", "Aaron Smith", "Bernd Bohnet", "Miryam de Lhoneux", "Joakim Nivre", "Yan Shao", "Sara Stymne", "Ömer Kırnap", "Erenay Dayanık", "Deniz Yuret", "Jenna Kanerva", "Filip Ginter", "Niko Miekka", "Akseli Leino", "Tapio Salakoski", "Kyungtae Lim", "Cheoneum Park", "Changki Lee", "Thierry Poibeau", "Riyaz Ahmad Bhat", "Irshad Bhat", "Srinivas Bangalore", "Peng Qi", "Timothy Dozat", "Yuhao Zhang", "Christopher Manning", "Tiberiu Boroş", "Stefan Daniel Dumitrescu", "Ruxandra Burtica", "Gor Arakelyan", "Karen Hambardzumyan", "Hrant Khachatrian", "Rudolf Rosa", "David Mareček", "Milan Straka", "Amit Seker", "Amir More", "Reut Tsarfaty", "Berkay Furkan Önder", "Can Gümeli", "Ganesh Jawahar", "Benjamin Muller", "Amal Fethi", "Louis Martin", "Eric Villemonte de la Clergerie", "Benoît Sagot", "Djamé Seddah", "Şaziye Betül Özateş", "Arzucan Özgür", "Tunga Gungor", "Balkız Öztürk", "Tao Ji", "Yufang Liu", "Yijun Wang", "Yuanbin Wu", "Man Lan", "Danlu Chen", "Mengxiao Lin", "Zhifeng Hu", "Xipeng Qiu" ]
https://www.semanticscholar.org/paper/fbd853ee77bf3c3e8ebbbaa3959f90a3e8cbb84a/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
CoNLL 2018 Shared Task: Multilingual Parsing from Raw Text to Universal Dependencies
Soutěž CoNLL 2018: Vícejazyčná syntaktická analýza z prostého textu do Universal Dependencies
Every year, the Conference on Computational Natural Language Learning (CoNLL) features a shared task, in which participants train and test their learning systems on the same data sets. In 2018, one of two tasks was devoted to learning dependency parsers for a large number of languages, in a real-world setting without any gold-standard annotation on test input.
Konference o počítačovém učení přirozeného jazyka (CoNLL) každoročně zahrnuje tzv. společnou úlohu, tedy soutěž, jejíž účastníci trénují a testují své systémy strojového učení na jednotné sadě dat. Jedna z úloh roku 2018 byla věnována učení závislostních syntaktických analyzátorů (parserů) pro velké množství jazyků, v reálné situaci bez jakýchkoli ručních anotací na vstupu.
[ "Daniel Zeman", "Jan Hajič", "Martin Popel", "Martin Potthast", "Milan Straka", "Filip Ginter", "Joakim Nivre", "Slav Petrov" ]
https://www.semanticscholar.org/paper/f0de0e88386eb49d770034c3f8663641549051be/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
ForFun 1.0: Prague Database of Forms and Functions -- An Invaluable Resource for Linguistic Research
ForFun 1.0: Pražská databáze forem a funkcí -- Neocenitelný zdroj pro lingvistický výzkum
NOT FINAL VERSION. In this paper, we introduce first version of ForFun, Prague Database of Syntactic Forms and Functions, as an invaluable resource for profound linguistic research, particularly in describing syntactic functions and their formal realizations. ForFun takes advantage of sev- eral richly syntactically annotated corpora, collectively called Prague Dependency Treebanks. The ForFun brings this rich and complex annotation of Czech sentences closer to common researchers. We demonstrate that the ForFun 1.0 provides valuable and rich material allowing to elaborate various syntactic issues in depth. We believe that nowadays when corpus linguistics differs from traditional linguis- tics in its insistence on a systematic study of authentic examples of language in use, our database will contribute to the comprehensive syntactic description.
V příspěvku představujeme první verzi Pražské databáze forem a funkcí, ForFun, jakožto neocenitelný zdroj jazykových dat pro lingvistický výzkum, zejména při popisu syntaktických funkcí a jejich formálních realizací v češtině. ForFun je vybudován na několika bohatě syntakticky anotovaných korpusecg z rodiny Pražských závislostních korpusů.
[ "Marie Mikulová", "Eduard Bejček" ]
https://www.semanticscholar.org/paper/901d0c8592c1e304613eeaef0fcd4d591616312d/
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
What Can We Find Out about Time and Space in the ForFun Database?
Co můžeme zjistit o čase a prostoru v databázi ForFun?
We present a description of time and space modifications in Czech sentences based on the ForFun database. ForFun is a new resource built on annotated corpora of Czech—Prague Dependency Treebanks—for inspecting thousands of real examples categorized by their form as well as by their deep syntactic function. Based on the database, we perform a detailed description of meanings of time and space modifications including a list of formal means with real examples coming from both written and spoken texts. It should be emphasized that the study is data-oriented rather than theory-oriented.
Předkládáme popis času a prostoru v českých větách na základě databáze ForFun. ForFun je nový lingvistický zdroj postavený na Pražskými závislostními korpusy češtiny. Umožňuje prohledávání tisíců skutečných příkladů tříděných podle užitých forem a také hloubkově syntaktických funkcí. Na základě databáze provedeme podrobný popis významů časových a prostorových určení včetně seznamu formálních prostředků. Příklady pocházejí z psaných i tištěných textů. Práce vychází z dat a neklade si za cíl nový teoretický popis.
[ "Marie Mikulová", "Eduard Bejček", "Jarmila Panevová" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
Search for the Relation of Form and Function Using the ForFun Database
Vyhledávání forem a jejich funkcí v datábázi ForFun
The aim of the contribution is to introduce a database of linguistic forms and their functions built with the use of the multi-layer annotated corpora of Czech, the Prague Dependency Treebanks. The purpose of the Prague Database of Forms and Functions (ForFun) is to help the linguists to study the form-function relation, which we assume to be one of the principal tasks of both theoretical linguistics and natural language processing. We demonstrate possibilities of the exploitation of the ForFun database. This article is largely based on a paper presented at the 16th International Workshop on Treebanks and Linguistic Theories in Prague (Bejček et al., 2017).
Představujeme databázi forem a funkcí vybudovanou na podkladě ručně anotovaných víceúrovňových korpusů češtiny zvaných Pražské závislostní korpusy. Pražská databáze forem a funkcí (ForFun) byla vytvořena, aby usnadnila lingvistické badání o vztahu funkce a formy, což je jeden z hlavních úkolů, jak v teoretické lingvistice, tak v oblasti počítačového zpracování jazyka. V článku jsou představeny možnosti, jakým způsobem lze v databázi vztah funkce a formy zkoumat. Článek je z větší části založen na již dříve prezentovaném příspěvku na 16th International Workshop on Treebanks and Linguistic Theories in Prague (Bejček et al., 2017).
[ "Marie Mikulová", "Eduard Bejček", "Eva Hajičová", "Jarmila Panevová" ]
https://www.semanticscholar.org/paper/4ddd7c8a68862face4e7989fb96ed4b63074eccf/
null
null
null
null
null
null
null
null
null
null
null
null
cs
2018
Between Reflexivity and Reciprocity: Towards an Analysis of Reflexive and Reciprocal Constructions of Selected Czech Verbs
Mezi reflexivitou a reciprocitou: Poznámky k reflexivním a recipročním konstrukcím vybraných českých sloves
In this article, we focus on two groups of Czech verbs, lexical and syntactic reciprocals. We provide an analysis of their syntactic properties with respect to reciprocity and reflexivity, their possible ambiguity and combination. We demonstrate that it is the reflexive pronoun and its surface expression that play a key role in the studied phenomena. In case of ambiguity, both reciprocity and reflexivity affect the same pair of situational participants / valency complementations; the reflexive pronoun occupies the same syntactic position and it corefers with the plural subject. In contrast, when these phenomena combine, different pairs of participants / complementations are involved. We also show that lexical and syntactic reciprocals differ with respect to the studied phenomena.
Na pozadí dvou skupin sloves, lexikálních a syntaktických reciprok v češtině, diskutujeme otázky homonymie a kombinovatelnosti reflexivity a reciprocity. Ukazujeme, že zdrojem homonymie i možnosti kombinace reflexivity a reciprocity je reflexivní zájmeno a že klíčovou úlohu sehrává jeho povrchověsyntaktická pozice. V případě homonymie je při reciprocitě a reflexivitě zasažena stejná dvojice situačních participantů / valenčních doplnění, reflexivní zájmeno tedy obsazuje totožnou povrchověsyntaktickou pozici a koreferuje s plurálovým subjektem. V případě kombinací těchto významů jde o rozdílné pozice a reflexivita a reciprocita zasahují různé dvojice participantů / doplnění. Uvedené dvě skupiny sloves se přitom vzhledem k homonymii a kombinovatelnosti těchto významů chovají do značné míry odlišně.
[ "Markéta Lopatková", "Václava Kettnerová" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
en
2018
The Tricky Case of Czech Reflexives
Problematická česká reflexiva
The talk focused on multiple functions which the reflexives have in the Czech language and their representation in a lexicon.
Příspěvek se zaměřil na polyfunkčnost reflexiv v češtině, vymezení jejich hlavní funckí, které v jazyce plní a důsledky pro jejich popis ve slovníku.
[ "Markéta Lopatková", "Václava Kettnerová" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
cs
2018
VALLEX 3.5 - Valency Lexicon of Czech Verbs
VALLEX 3.5 - Valenční slovník českých sloves
VALLEX 3.5 provides information on the valency structure (combinatorial potential) of verbs in their particular senses, which are characterized by glosses and examples. VALLEX 3.5 describes almost 4 600 Czech verbs in more than 10 700 lexical units, i.e., given verbs in the given senses; this version is enriched with an annotation of light verb constructions -- almost 1 500 complex predicates have been added. VALLEX is a is a collection of linguistically annotated data and documentation, resulting from an attempt at formal description of valency frames of Czech verbs. In order to satisfy different needs of different potential users, the lexicon is distributed (i) in a HTML version (the data allows for an easy and fast navigation through the lexicon) and (ii) in a machine-tractable form as a single XML file, so that the VALLEX data can be used in NLP applications.
Valenční slovník českých sloves (též VALLEX 3.5) poskytuje informace o valenční struktuře českých sloves v jejich jednotlivých významech, které charakterizuje pomocí glos a příkladů; tyto základní údaje jsou doplněny o některé další syntaktické a sémantické charakteristiky. VALLEX 3.5 zachycuje téměř 4 600 českých sloves, která odpovídají více než 10 700 lexikálním jednotkám, tedy vždy danému slovesu v daném významu; tato verze je rozšířena o anotaci komplexních predikátů s kategoriálním slovesem (zpracováno celkem téměř 1 500 komplexních predikátů). VALLEX je budován jako odpověď na potřebu rozsáhlého a kvalitního veřejně dostupného slovníku, který by obsahoval syntaktické a sémantické charakteristiky českých sloves a byl přitom široce uplatnitelný v aplikacích pro zpracování přirozeného jazyka. Při navrhování koncepce slovníku byl proto kladen důraz na všestranné využití jak pro lidského uživatele, tak pro automatické zpracování češtiny.
[ "Markéta Lopatková", "Václava Kettnerová", "Eduard Bejček", "Anna Vernerová", "Zdeněk Žabokrtský", "Petra Barančíková" ]
https://www.semanticscholar.org/paper/975e45e9396bcdd6f76be4f89deccd900b57613f/
null
null
null
null
null
null
null
null
null
null
null
null
cs
2018
DATLOWE: software for text analytics
DATLOWE: software pro textovou analytiku
Licence and service agreement between Charles University and Datlowe Ltd.(January 1, 2018 -- January 1, 2020) covers the right to use the data of PDT 2.5 and the software for text analytics trained on PDT 2.5, for commercial purposes. The data and software (non-adapted) is otherwise available for free through the LINDAT/CLARIN repository for non-commercial use under the appropriate variant of the Creative Commons license. Namely, software for morphological analysis, lemmatization, syntactic analysis and named-entity recognition.
Licenční smlouva a smlouva o poskytnutí služeb mezi Univerzitou Karlovou a Datlowe, s.r.o. (1. 1. 2018 -- 1. 1. 2020) obsahuje právo na užití dat PDT 2.5 a software pro textovou analytiku natrénovaného na PDT 2.5, a to pro komerční účely. Jedná se o upravené verze nekomerčního software a dat jinak dostupných z repozitáře LINDAT/CLARIN. Jmenovitě: software pro analýzu tvarosloví a lematizaci, syntaktickou analýzu a analýzu názvů pojmenovanýchch entit.
[ "Markéta Lopatková", "Barbora Hladká" ]
null
null
null
null
null
null
null
null
null
null
null
null
null