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
2012
TrTok: A Fast and Trainable Tokenizer for Natural Languages
TrTok: Rychlý a trénovatelný tokenizér pro přirozené jazyky
We present a universal data-driven tool for segmenting and tokenizing text. The presented tokenizer lets the user define where token and sentence boundaries should be considered. These instances are then judged by a classifier which is trained from provided tokenized data. The features passed to the classifier are also defined by the user making, e.g., the inclusion of abbreviation lists trivial. This level of customizability makes the tokenizer a versatile tool which we show is capable of sentence detection in English text as well as word segmentation in Chinese text. In the case of English sentence detection, the system outperforms previous methods. The software is available as an open-source project on GitHub
Představujeme univerzální nástroj pro segmentaci a tokenizaci textů, který uživateli dovoluje nadefinovat potenciální hranice vět a slov a na základě trénovacích dat se naučí hranice hledat.
[ "Jiří Maršík", "Ondřej Bojar" ]
https://www.semanticscholar.org/paper/875263b7ae4749a017b8a008d261096e6458c556/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
Building parallel corpora through social network gaming
Budování paralelních korpusů pomocí her na sociálních sítích
Building training data is labor-intensive and presents a major obstacle to the advancement of Natural Language Processing (NLP) systems. A prime use of NLP technologies has been toward the construction machine translation systems. The most common form of machine translation systems are phrase based systems that require extensive training data. Building this training data is both expensive and error prone. Emerging technologies, such as social networks and serious games, offer a unique opportunity to change how we construct training data. These serious games, or games with a purpose, have been constructed for sentence segmentation, image labeling, and co-reference resolution. These games work on three levels: They provide entertainment to the players, the reinforce information the player might be learning, and they provide data to researchers. Most of these systems while well intended and well developed, have lacked participation. We present, a set of linguistically based games that aim to construct parallel corpora for a multitude of languages and allow players to start learning and improving their own vocabulary in these languages. As of the first release of the games, GlobeOtter is available on Facebook as a social network game. The release of this game is meant to change the default position in the field, from creating games that only linguists play, to releasing linguistic games on a platform that has a natural user base and ability to grow.
Příprava trénovacích dat je pracná a představuje hlavní překážku ve vývoji zpracování přirozeného jazyka (NLP). Mezi hlavní aplikace NLP patří strojový překlad, který se v současnosti opírá o dostupnost velkého množství dat. Sestavování těchto dat je finančně náročné a současně náchylné k chybám. Nově se objevující technologie jako sociální sítě a "seriózní" hry nabízejí jedinečnou příležitost změnit způsob přípravy trénovacích dat. Hry s účelem byly zkonstruovány pro větnou segmentaci, značkování obrázků a rozpoznávání koreference. Tyto hry fungují na třech úrovních: poskytují zábavu hráčům, hráči se při nich učí a současně poskytují data pro výzkum. Většina těchto systémů se potýká s nedostatkem účastníků. V tomto článku předkládáme sadu lingvisticky orientovaných her zaměřených na sestrojení paralelního korpusu pro několik jazyků a umožňujících hráčům zlepšení jejich slovní zásoby v těchto jazycích. První zveřejněná verze GlobeOtter je dostupná na Facebooku. Jedním z cílů je zde získat dostatečné množství hráčů i mimo řady lingvistů.
[ "Nathan David Green" ]
https://www.semanticscholar.org/paper/58a2eb7108ff43d821183dd52bfb4030c5a6132d/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
Indonesian Dependency Treebank: Annotation and Parsing
Závislostní korpus indonésštiny: anotace a parsing
We also show ensemble dependency parsing and self training approaches applicable to under-resourced languages using our manually annotated dependency structures. We show that for an under-resourced language, the use of tuning data for a meta classifier is more effective than using it as additional training data for individual parsers. This meta-classifier creates an ensemble dependency parser and increases the dependency accuracy by 4.92% on average and 1.99% over the best individual models on average. As the data sizes grow for the the under-resourced language a meta classifier can easily adapt. To the best of our knowledge this is the first full implementation of a dependency parser for Indonesian. Using self-training in combination with our Ensemble SVM Parser we show additional improvement. Using this parsing model we plan on expanding the size of the corpus by using a semi-supervised approach by applying the parser and correcting the errors, reducing the amount of annotation time needed.
Článek ukazuje závislostní analýzu s využitím kombinace parseru a self-trainingu pro jazyky s malým množstvím jazykových dat. Ověřili jsme, že pro jazyk s malým množstvím dat je využití ladicích dat pro meta-klasifikátor efektivnější než jejich přidání do zbývajících trénovacích dat jednotlivých analyzátorů. Tento mete-klasifikátor vytváří kombinovaný závislostní parse a zvyšuje úspěšnost analýzy v průměru o 4.92% a o 1.99% ve srovnání s jednotlivým nejlepším systémem. Meta-klasifikátor se může přizpůsobit rostoucím dostupným datům. Využitím self-trainingu společně s kombinací několika parseru vzniká další zlepšení.
[ "Nathan David Green", "Septina Dian Larasati", "Zdeněk Žabokrtský" ]
https://www.semanticscholar.org/paper/39d6db85b34319e9327ef69b0ec1fcdf95d93b32/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
Ensemble Parsing and its Effect on Machine Translation
Kombinace syntaktických analyzátorů a její vliv na strojový překlad
The focus of much of dependency parsing is on creating new modeling techniques and examining new feature sets for existing dependency models. Often these new models are lucky to achieve equivalent results with the current state of the art results and often perform worse. These approaches are for languages that are often resource-rich and have ample training data available for dependency parsing. For this reason, the accuracy scores are often quite high. This, by its very nature, makes it quite difficult to create a significantly large increase in the current state-of-the-art. Research in this area is often concerned with small accuracy changes or very specific localized changes, such as increasing accuracy of a particular linguistic construction. With so many modeling techniques available to languages with large resources the problem exists on how to exploit the current techniques with the use of combination, or ensemble, techniques along with this plethora of data. Dependency parsers are almost ubiquitously evaluated on their accuracy scores, these scores say nothing of the complexity and usefulness of the resulting structures. The structures may have more complexity due to the depth of their co- ordination or noun phrases. As dependency parses are basic structures in which other systems are built upon, it would seem more reasonable to judge these parsers down the NLP pipeline. The types of parsing errors that cause significant problems in other NLP applications is currently an unknown.
Fokus hodně závislostní analýzy se na vytváření nových modelovacích technik a zkoumání nových funkcí sady pro stávající závislost modelů. Často jsou tyto nové modely mají to štěstí, aby dosahovaly rovnocenných výsledků s aktuálním stavem že umělecké výsledky a často vedou hůř. Tyto přístupy jsou pro jazyky, které jsou často zdrojem bohaté a mají dostatek tréninková data k dispozici pro závislost rozebrat. Z tohoto důvodu, přesnost výsledky jsou často dosti vysoká. To, ze své podstaty, je to docela obtížné vytvořit výrazně velký nárůst v současném state-of-the-art. Výzkum v této oblasti se často zabývá malými změnami přesnosti nebo velmi specifickým lokalizovaných změn, jako je zvýšení přesnosti a zejména jazykovou konstrukci. S tolika technik modelování jsou k dispozici pro jazyky s velkým zdroji problém existuje o tom, jak využít stávající techniky s použitím kombinace, nebo souborem, techniky spolu s tímto množstvím dat. Závislost analyzátory jsou téměř všude se vyskytující hodnoceny na jejich přesnost výsledků, tyto výsledky nemluvě o složitosti a užitečnosti výsledných struktur. Tyto struktury mohou mít větší složitost vzhledem k hloubce jejich co- koordinační nebo podstatné jméno věty. Jako závislost analyzuje jsou základní struktury, ve kterých jsou ostatní systémy postavené na, by se zdálo rozumné posouzení těchto parserů dolů potrubí NLP. Typy parsování chyb, které způsobují významné problémy v jiných aplikacích NLP je v současné době znám.
[ "Nathan David Green", "Zdeněk Žabokrtský" ]
https://www.semanticscholar.org/paper/7f37d7a115fdf55b27fa954adfe6e66fe22ea4cd/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
Hybrid Combination of Constituency and Dependency Trees into an Ensemble Dependency Parser
Hybridní kombinace složkových a závislostních stromů v kombinačním parseru
Dependency parsing has made many advancements in recent years, in particular for English. There are a few dependency parsers that achieve comparable accuracy scores with each other but with very different types of errors. This paper examines creating a new dependency structure through ensemble learning using a hybrid of the outputs of various parsers. We combine all tree outputs into a weighted edge graph, using 4 weighting mechanisms. The weighted edge graph is the input into our ensemble system and is a hybrid of very different parsing techniques (constituent parsers, transition-based dependency parsers, and a graph-based parser). From this graph we take a maximum spanning tree. We examine the new dependency structure in terms of accuracy and errors on individual part-of-speech values. The results indicate that using a greater number of more varied parsers will improve accuracy results. The combined ensemble system, using 5 parsers based on 3 different parsing techniques, achieves an accuracy score of 92.58%, beating all single parsers on the Wall Street Journal section 23 test set. Additionally, the ensemble system reduces the average relative error on selected POS tags by 9.82%.
V posledních letech dosáhla závislostní syntaktická analýza podstatného zlepšení, zejména pro angličtnu. Existuje několik závislostních analyzátorů, které dosahují srovnatelné úspěšnosti, přičemž produkují různé typy chyb. Tento článek se zabývá vytvářením nové závislostní struktury složením výstupů různých analyzátorů. Všechny výstupy sloučíme do váženého grafu, který je vstupem pro kombinační systém založený na algoritmu minimální kostry orientovaného grafu. Výsledná kombinace založená na pěti různých analyzátorech dosahuje úspěšnosti 92,58 %, čímž překonává všechny jednotlivé analyzátory.
[ "Nathan David Green", "Zdeněk Žabokrtský" ]
https://www.semanticscholar.org/paper/b4c3b21c4048e2a001f5a698dc8376054be64145/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
Using an SVM Ensemble System for Improved Tamil Dependency Parsing
Klasifikace pomocí SVM v kombinačním závislostním analyzátoru tamilštiny
Dependency parsing has been shown to improve NLP systems in certain languages and in many cases helps achieve state of the art results in NLP applications, in particular applications for free word order languages. Morphologically rich languages are often short on training data or require much higher amounts of training data due to the increased size of their lexicon. This paper examines a new approach for addressing morphologically rich languages with little training data to start. Using Tamil as our test language, we create 9 dependency parse models with a limited amount of training data. Using these models we train an SVM classifier using only the model agreements as features. We use this SVM classifier on an edge by edge decision to form an ensemble parse tree. Using only model agreements as features allows this method to remain language independent and applicable to a wide range of morphologically rich languages. We show a statistically significant 5.44% improvement over the average dependency model and a statistically significant 0.52% improvement over the best individual system.
Závislostní syntaktická analýza pomáhá některým aplikacím z oblasti zpracování přirozeného jazyka dosáhnout vyšší úspěšnosti, zejména pokud jde o jazyky s relativně volným slovosledem. Pro morfologicky bohaté jazyky typicky existuje jen malé množství trénovacích dat, přičemž kvůli větší velikosti slovníku by jich naopak bylo potřeba více. Tento článek se zabývá novými přístupy pro analýzu morfologicky bohatých jazyků s malým množstvím dat. Testovacím jazykem je v našich experimentech tamilština. Vytvořili jsme 9 modelů pro závislostní syntaktickou analýzu, které byly natrénovány na malém množství dat. S využitím těchto modelů jsme natrénovali klasifikátor SVM, který jako rysy používá pouze informaci o shodě jednotlivých analyzátorů, díky čemuž lze tento přístup považovat za jazykově nezávislý. Experimentálně jsme prokázali statisticky signifikantní zlepšení 5,44 % oproti průměrnému modelu a statisticky signifikantní zlepšení 0,52 % oproti nejlepšímu jednotlivému systému.
[ "Nathan David Green", "Loganathan Ramasamy", "Zdeněk Žabokrtský" ]
https://www.semanticscholar.org/paper/09dd68be32eea757d509ed359932df7990308c5d/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
Application of Topic Segmentation in Audiovisual Information Retrieval
Využitie tématickej segmentácie pri vyhľadávaní v audiovizuálnych dátach
Segmentation into topically coherent segments is one of the crucial points in information retrieval (IR). Suitable segmentation may improve the results of IR system and help users to find relevant passages faster. Segmentation is especially important in audiovisual recordings, in which the navigation is difficult. We present several methods used for topic segmentation, based on textual, audio and visual information. The proposition of our approach to topic segmentation based on the fusion of audio and visual data is presented in the article.
Segmentácia na tématicky koherentné úseky je dôležitou súčasťou vyhľadávania informácií. Vhodná segmentácia môže zlepšiť výsledky vyhľadávania a pomôcť užívateľom pri rýchlejšom hľadaní relevantného úseku. Segmentácia je zvlášť dôležitá pri audiovizuálnych nahrávkach, v ktorých je navigácia zvlášť zložitá. V článku popisujeme niekoľko prístupov ku tématickej segmentácii, založených na textovej, zvukovej a vizuálnej informácii. V článku je tiež prezentovaný náš návrh prístupu k tématickej segmentácii, založený na fúzii zvukových a vizuálnych dát.
[ "Petra Galuščáková" ]
https://www.semanticscholar.org/paper/654c57b29698c7a14959caf0e996c7aeb3eb0e3e/
null
null
null
null
null
null
null
null
null
null
null
null
cs
2012
Czech-Slovak Parallel Corpus
Česko-slovenský paralelný korpus
Czech-Slovak parallel corpus consisting of several freely available corpora. Corpus is given in both plaintext format and with an automatic morphological annotation.
Česko-slovenský paralelný korpus, vytvorený z voľne dostupných zdrojov. Korpus je prístupný v textovom formáte a s morfologickou anotáciou.
[ "Petra Galuščáková", "Radovan Garabík", "Ondřej Bojar" ]
https://www.semanticscholar.org/paper/eef4d17f0977d590535546310d2e4a6852bd3747/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
English-Slovak Parallel Corpus
Anglicko-slovenský paralelný korpus
English-Slovak parallel corpus consisting of several freely available corpora. Corpus is given in both in plaintext format and with an automatic morphological annotation
Anglicko-slovenský paralelný korpus, vytvorený z voľne dostupných zdrojov. Korpus je prístupný v textovom formáte a s morfologickou anotáciou.
[ "Petra Galuščáková", "Radovan Garabík", "Ondřej Bojar" ]
https://www.semanticscholar.org/paper/0c42a29ef3e802a7ccdbe589cdce2013ca82039f/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
Improving SMT by Using Parallel Data of a Closely Related Language
Vylepšenie štatistického automatického prekladu pomocou paralelných dát blízkych jazykov
The amount of training data in statistical machine translation is critical for translation quality. In this paper, we demonstrate how to increase translation quality for one language pair by bringing in parallel data from a closely related language. In particular, we improve en→sk translation using a large Czech–English parallel corpus and a shallow (rule-based) MT system for cs→sk. Several setup options are explored in order to identify the best possible configuration.
Množstvo trénovacích dát je pre kvalitu štatistického strojového prekladu rozhodujúce. V článku popisujeme, akým spôsobom je možné zlepšiť kvalitu prekladu pre daný jazykový pár pomocou využitia paralelných dát v príbuznom jazyku. Konkrétne sme vylepšili en→sk preklad pomocou využitia veľkého česko-anglického paralelného korpusu a cs→sk prekladového systému založeného na pravidlách. Preskúmaných je niekoľko možností konfigurácie použitých systémov.
[ "Petra Galuščáková", "Ondřej Bojar" ]
https://www.semanticscholar.org/paper/8f270392dbf7bf35e2cc05b1f51e130521ace9a4/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
Manually Classified Errors in Czech-Slovak Translation
Ručne klasifikované chyby v česko-slovenskom preklade
Outputs of five Czech-Slovak machine translation systems (Česílko, Česílko 2, Google Translate and Moses with different settings) for first 50 sentences of WMT 2010 testing set. The translations were manually processed and the errors were marked and classified according to the scheme by Vilar et al. (David Vilar, Jia Xu, Luis Fernando D’Haro, Hermann Ney: Error Analysis of Statistical Machine Translation Output, Proceedings of LREC-2006, 2006)
Automatický preklad z češtiny do slovenčiny pre prvých 50 viet testovacej sady WMT 2010 pomocou piatich prekladových systémov (Česílko, Česílko 2, Google Translate a Moses s rôznymi nastaveniami). V prekladoch boli ručne označené chyby, ktoré boli ďalej klasifikované podľa schémy danej v článku (David Vilar, Jia Xu, Luis Fernando D’Haro, Hermann Ney: Error Analysis of Statistical Machine Translation Output, Proceedings of LREC-2006, 2006).
[ "Petra Galuščáková", "Ondřej Bojar" ]
https://www.semanticscholar.org/paper/7376e7d2228ae760dcf37670fe97d01a6b161d35/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
Manually Classified Errors in English-Slovak Translation
Ručne klasifikované chyby v anglicko-slovenskom preklade
Outputs of three English-Slovak machine translation systems for 50 sentences randomly selected from WMT 2011 test set. The translations were manually processed and the errors were marked and classified according to the scheme by Vilar et al. (David Vilar, Jia Xu, Luis Fernando D’Haro, Hermann Ney: Error Analysis of Statistical Machine Translation Output, Proceedings of LREC-2006, 2006)
Automatický preklad z angličtiny do slovenčiny 50 viet náhodne vybraných z testovacej sady WMT 2011 pomocou troch prekladových systémov. V prekladoch boli ručne označené chyby, ktoré boli ďalej klasifikované podľa schémy danej v článku (David Vilar, Jia Xu, Luis Fernando D’Haro, Hermann Ney: Error Analysis of Statistical Machine Translation Output, Proceedings of LREC-2006, 2006).
[ "Petra Galuščáková", "Ondřej Bojar" ]
https://www.semanticscholar.org/paper/7376e7d2228ae760dcf37670fe97d01a6b161d35/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
WMT 2011 Testing Set in Slovak
Testovacia sada WMT 2011 preložená do slovenčiny
WMT 2011 Workshop testing set manually translated from Czech and English into Slovak.
Testovacia sada z workshopu WMT 2011 manuálne preložená z češtiny a angličtiny do slovenčiny.
[ "Petra Galuščáková", "Ondřej Bojar" ]
https://www.semanticscholar.org/paper/d04740ce1d32df9d3b80a2ae9ee81ab3aa7c11f6/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
CUNI at MediaEval 2012 Search and Hyperlinking Task
CUNI na MediaEvale 2012, Search and Hyperlinking Task
The paper describes the Charles University setup used in the Search and Hyperlinking task of the MediaEval 2012 Multimedia Benchmark. We applied the Terrier retrieval system to the automatic transcriptions of the video recordings segmented into shorter parts and searched for those relevant to given queries. Two strategies were applied for segmentation of the recordings: one based on regular segmentation according to time and the second based on semantic segmentation by the TextTiling algorithm. The best results were achieved by the Hiemstra and TF-IDF models on the LIMSI transcripts and various segmentation.
Článok popisuje postup, ktorý sme použili v úlohe Search and Hyperlinking v MediaEval 2012 Multimedia Benchmark. Pri riešení bol použitý systém Terrier, ktorý bol aplikovaný na automatické prepisy video nahrávok. Nahrávky boli segmentované na menšie časti, ktoré boli ďalej prehľadané. Pri segmentácii boli použité dva prístupy: jeden založený na pravidelnej segmentácii podľa času a druhý založený na sémantickej segmentácii pomocou algoritmu TextTiling. Najlepší výsledok bol dosiahnutý pomocou pravidelnej segmentácie, pomocou modelov Hiemstra a TF-IDF na prepisoch LIMSI.
[ "Petra Galuščáková", "Pavel Pecina" ]
https://www.semanticscholar.org/paper/1834ffde41d84bd9bfcd063942a9936433723191/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
Penalty Functions for Evaluation Measures of Unsegmented Speech Retrieval
Penalizačné funkcie v metrikách na vyhodnocovanie vyhľadávania v hovorenej reči
This paper deals with evaluation of information retrieval from unsegmented speech. We focus on Mean Generalized Average Precision, the evaluation measure widely used for unsegmented speech retrieval. This measure is designed to allow certain tolerance in matching retrieval results (starting points of relevant segments) against a gold standard relevance assessment. It employs a Penalty Function which evaluates non-exact matches in the retrieval results based on their distance from the beginnings of their nearest true relevant segments. However, the choice of the Penalty Function is usually ad-hoc and does not necessary reflect users’ perception of the speech retrieval quality. We perform a lab test to study satisfaction of users of a speech retrieval system to empirically estimate the optimal shape of the Penalty Function.
Článok sa zaoberá vyhodnocovaním vyhľadávania v nesegmentovanej hovorenej reči. Zameriavame sa na metriku Mean Generalized Average precision, ktorá sa vyhodnocovanie vyhľadávania v reči často využíva. Táto metrika je navrhnutá tak, aby tolerovala určitú odchýlku medzi automaticky vyhľadanými výsledkami (začiatočnými bodmi relevantných segmentov) a ručne označenými výsledkami. Na tento účel používa metrika penalizačnú funkciu, ktorá určuje kvalitu vyhľadaných úsekov na základe ich vzdialenosti ku najbližšiemu začiatku ručne označeného relevantného úseku. Výber penalizačnej funkcie však nemusí zodpovedať kvalite vyhľadávania z pohľadu užívateľa. V článku popisujeme užívateľskú štúdiu, ktorou sme skúmali spokojnosť užívateľa pri vyhľadávaní informácií. Na základe výsledkov sme navrhli optimálnu penalizačnú funkciu.
[ "Petra Galuščáková", "Pavel Pecina", "Jan Hajič" ]
https://www.semanticscholar.org/paper/57a3f36711ca254cd6bb4d702f29be5fe62f177c/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
A Dataset Comparison for an Indonesian-English Statistical Machine Translation System
N/A
In this paper, we study the effect of incorporating morphological information on an Indonesian (id) to English (en) Statistical Machine Translation (SMT) system as part of a preprocessing module. The linguistic phenomenon that is being addressed here is Indonesian cliticized words. The approach is to transform the text by separating the correct clitics from a cliticized word to simplify the word alignment. We also study the effect of applying the preprocessing on different SMT systems trained on different kinds of text, such as spoken language text. The system is built using the state-of-the-art SMT tool, MOSES. The Indonesian morphological information is provided by MorphInd. Overall the preprocessing improves the translation quality, especially for the Indonesian spoken language text, where it gains 1.78 BLEU score points of increase.
(Not yet available. English version repeated) In this paper, we study the effect of incorporating morphological information on an Indonesian (id) to English (en) Statistical Machine Translation (SMT) system as part of a preprocessing module. The linguistic phenomenon that is being addressed here is Indonesian cliticized words. The approach is to transform the text by separating the correct clitics from a cliticized word to simplify the word alignment. We also study the effect of applying the preprocessing on different SMT systems trained on different kinds of text, such as spoken language text. The system is built using the state-of-the-art SMT tool, MOSES. The Indonesian morphological information is provided by MorphInd. Overall the preprocessing improves the translation quality, especially for the Indonesian spoken language text, where it gains 1.78 BLEU score points of increase.
[ "Septina Dian Larasati" ]
https://www.semanticscholar.org/paper/29cf82bfd225650d5c42e2ace729b662a94cf422/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
IDENTIC Corpus: Morphologically Enriched Indonesian-English Parallel Corpus
IDENTIC Corpus: Indonésko-anglický paralelní korpus obohacený o morfologii
This paper describes the creation process of an Indonesian-English parallel corpus (IDENTIC). The corpus contains 45,000 sentences collected from different sources in different genres. Several manual text preprocessing tasks, such as alignment and spelling correction, are applied to the corpus to assure its quality. We also apply language specific text processing such as tokenization on both sides and clitic normalization on the Indonesian side. The corpus is available in two different formats: ‘plain’, stored in text format and ‘morphologically enriched’, stored in CoNLL format. Some parts of the corpus are publicly available at the IDENTIC homepage.
This paper describes the creation process of an Indonesian-English parallel corpus (IDENTIC). The corpus contains 45,000 sentences collected from different sources in different genres. Several manual text preprocessing tasks, such as alignment and spelling correction, are applied to the corpus to assure its quality. We also apply language specific text processing such as tokenization on both sides and clitic normalization on the Indonesian side. The corpus is available in two different formats: ‘plain’, stored in text format and ‘morphologically enriched’, stored in CoNLL format. Some parts of the corpus are publicly available at the IDENTIC homepage
[ "Septina Dian Larasati" ]
https://www.semanticscholar.org/paper/c638af7a620b30459db2e488a99f86bc50d486a8/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
Improving Word Alignment by Exploiting Adapted Word Similarity
N/A
This paper presents a method to improve a word alignment model in a phrase-based Statistical Machine Translation system for a low resourced language using a string similarity approach. Our method captures similar words that can be seen as semi-monolingual across languages, such as numbers, named entities, and adapted/loan words. We use several string similarity metrics to measure the monolinguality of the words, such as Longest Common Subsequence Ratio (LCSR), Minimum Edit Distance Ratio (MEDR), and we also use a modified BLEU Score (modBLEU). Our approach is to add intersecting alignment points for word pairs that are orthographically similar, before applying a word alignment heuristic, to generate a better word alignment. We demonstrate this approach on Indonesian-to-English translation task, where the languages share many similar words that are poorly aligned given a limited training data. This approach gives a statistically significant improvement by up to 0.66 in terms of BLEU score.
(Not yet available. English version repeated) This paper presents a method to improve a word alignment model in a phrase-based Statistical Machine Translation system for a low resourced language using a string similarity approach. Our method captures similar words that can be seen as semi-monolingual across languages, such as numbers, named entities, and adapted/loan words. We use several string similarity metrics to measure the monolinguality of the words, such as Longest Common Subsequence Ratio (LCSR), Minimum Edit Distance Ratio (MEDR), and we also use a modified BLEU Score (modBLEU). Our approach is to add intersecting alignment points for word pairs that are orthographically similar, before applying a word alignment heuristic, to generate a better word alignment. We demonstrate this approach on Indonesian-to-English translation task, where the languages share many similar words that are poorly aligned given a limited training data. This approach gives a statistically significant improvement by up to 0.66 in terms of BLEU score.
[ "Septina Dian Larasati" ]
https://www.semanticscholar.org/paper/91571c7aa106ea2b11e1e8998b007a3b6fac4b66/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
Towards an Indonesian-English SMT System: A Case Study of an Under-Studied and Under-Resourced Language, Indonesian
N/A
This paper describes a work on preparing an Indonesian-English Statistical Machine Translation (SMT) System. It includes the creation of Indonesian morphological analyzer, MorphInd, and the composing of an Indonesian-English parallel corpus, IDENTIC. We build an SMT system using the state-of-the-art phrase-based SMT system, MOSES. We show several scenarios where the morphological tool is used to incorporate morphological information in the SMT system trained with the composed parallel corpus.
(Not yet available. English version repeated) This paper describes a work on preparing an Indonesian-English Statistical Machine Translation (SMT) System. It includes the creation of Indonesian morphological analyzer, MorphInd, and the composing of an Indonesian-English parallel corpus, IDENTIC. We build an SMT system using the state-of-the-art phrase-based SMT system, MOSES. We show several scenarios where the morphological tool is used to incorporate morphological information in the SMT system trained with the composed parallel corpus.
[ "Septina Dian Larasati" ]
https://www.semanticscholar.org/paper/47e1a2c4d950d68d2ac7a69c155831edac242d9f/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
Tagger Voting for Urdu
Hlasování taggerů pro urdštinu
V článku popisujeme, jak jsme zlepšili značkování slovních druhů pro urdštinu pomocí kombinace dostupných taggerů a dostupných ručně značkovaných dat. V prvním kroku sjednocujeme sady značek užívané v jednotlivých zdrojích. Dále náš vlastní tagger natrénovaný na dostupných datech funguje výrazně lépe než dostupné nástroje. A konečně tento výsledek je možné ještě mírně zlepšit za použití návrhů od ostatních taggerů.
In this paper, we focus on improving part-of-speech (POS) tagging for Urdu by using existing tools and data for the language. In our experiments, we use Humayoun’s morphological analyzer, the POS tagging module of an Urdu Shallow Parser and our own SVM Tool tagger trained on CRULP manually annotated data. We convert the output of the taggers to a common format and more importantly unify their tagsets. On an independent test set, our tagger outperforms the other tools by far. We gain some further improvement by implementing a voting strategy that allows us to consider not only our tagger but also include suggestions by the other tools. The final tagger reaches the accuracy of 87.98%.
[ "Bushra Jawaid", "Ondřej Bojar" ]
https://www.semanticscholar.org/paper/572c7f65899b6934b813e7aa1a9c8f4bbcd3c54d/
null
null
null
null
null
null
null
null
null
null
null
null
cs
2012
Videotaped Witness Testimonies from the USC Shoah Foundation Institute's Visual History Archive as Educational Resource
Vzdělávací využití videozáznamů rozhovorů s pamětníky z Archivu vizuální historie Institutu USC Shoah Foundation
Witness testimonies are unsubstitutable and very important addition to the classic lessons of history and other subjects in primary and secondary education. The testimony allows a personalization of the „great history“, students can easily identify with a specific person and his/her experience, and it brings emotions into the educational process, which demonstrably improves its efficiency. However, the day will come, that there will be no more witnesses alive for certain historical events. It is thus necessary to explore and create new alternatives, that could – at least partially – substitute the presence of an eye-witness in the future classroom. The USC Shoah Foundation Institute Visual History Archive is a unique collection of almost 52 000 testimonies of the Holocaust survivors and witnesses, conducted in 56 countries and 32 languages during the late 90's (over 1 000 interviews are in Czech and Slovak language). The Visual History Archive provides complex tools for users to identify whole testimonies of relevance, as well as specific segments within testimonies that relate to their area of interest. Prague is currently one of five European cities with full access to the on-line licensed archive content from the Malach Visual History Centre at the Faculty of Mathematics and Physics of the Charles University. Real-time work with the archive is a valuable part of educational process as well, developing student's inde­pendence and critical thinking in context. This is a main purpose of the new educational portal IWitness, which is also going to be presented at the conference.
Svědectví pamětníka je nezastupitelným a velice hodnotným doplňkem běžné výuky dějepisu a dalších předmětů na 2. stupni ZŠ i na různých typech středních škol. Svědectví umožňuje personalizaci „velké historie“, žáci a studenti se s konkrétním člověkem a jeho prožitky mohou lépe identifikovat, do vzdělávacího procesu vstupují emoce, což prokazatelně zvyšuje jeho účinnost. Nevyhnutelně se však blíží doba, kdy už nebude možné zvát pamětníky přímo do škol a dalších institucí – už dnes zbývá jen velice málo lidí, kteří v dospělém věku prožili 2. světovou válku a události s ní spjaté. Je tedy třeba zjišťovat možnosti a metody alternativních postupů, které mohou alespoň částečně přítomnost živého pamětníka ve třídě v budoucnu nahradit. Archiv vizuální historie Institutu USC Shoah Foundation je unikátní sbírkou takřka 52 000 audiovi­zuálních záznamů rozhovorů s pamětníky a přeživšími holocaustu, které byly natočeny během 2. poloviny 90. let v 56 zemích a 32 jazycích (v češtině a slovenštině je přes 1 000 rozhovorů). Archiv je možno prohledávat pomocí propracovaného uživatelského rozhraní: uživatelé mohou najít konkrétní úseky svědectví podle svého zájmu díky tezauru 55 000 hierarchicky uspořádaných klíčových slov, témat, událostí, jmen osob, míst atd. Praha je jedním z pěti evropských měst, kde se lze k obsahu licencovaného archivu připojit, a to z počítačů v Centru vizuální historie Malach při Matematicko-fyzikální fakultě Univerzity Karlovy. Také práce s podobným on-line archivem v reálném čase je dalším specifickým vzdělávacím postupem, který u žáků rozvíjí samostatnost, myšlení v souvislostech a kritický přístup k pramenům a informačním zdrojům. Za tímto účelem vznikl také nový edukační portál IWitness, který bude v příspěvku rovněž představen.
[ "Jakub Mlynář" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
Khresmoi: Multimodal Multilingual Medical Information Search
Khresmoi: Multimodální a multilingvální hledání medicínských informací
Khresmoi is a European Integrated Project developing a multilingual multimodal search and access system for medical and health information and documents. It addresses the challenges of searching through huge amounts of medical data, including general medical information available on the internet, as well as radiology data in hospital archives. It is developing novel semantic search and visual search techniques for the medical domain. At the MIE Village of the Future, Khresmoi proposes to have two interactive demonstrations of the system under development, as well as an overview oral presentation and potentially some poster presentations.
Khresmoi je integrační projekt financovaný Evropskou Unií, který se zabývá multilingválním a multimodálním vyhledávání v medicínských datech.
[ "Niraj Aswani", "Thomas Beckers", "Erich Birngruber", "Célia Boyer", "Andreas Burner", "Jakub Bystroň", "Khalid Choukri", "Sarah Cruchet", "Hamish Cunningham", "Jan Dědek", "Ljiljana Dolamic", "René Donner", "Sebastian Dungs", "Ivan Eggel", "Antonio Foncubierta", "Norbert Fuhr", "Adam Funk", "Alba García Seco de Herrera", "Arnaud Gaudinat", "Georgi Georgiev", "Julien Gobeill", "Lorraine Goeuriot", "Paz Gomez", "Mark A. Greenwood", "Manfred Gschwandtner", "Allan Hanbury", "Jan Hajič", "Jaroslava Hlaváčová", "Markus Holzer", "Gareth J.F. Jones", "Blanca Jordán", "Matthias Jordan", "Klemens Kaderk", "Franz Kainberger", "Liadh Kelly", "Sascha Mriewel", "Marlene Kritz", "Georg Langs", "Nolan Lawson", "Dimitrios Markonis", "Iván Martínez", "Vassil Momtchev", "Alexandre Masselot", "Hélène Mazo", "Henning Müller", "Pavel Pecina", "Konstantin Pentchev", "Deyan Peychev", "Natalia Pletneva", "Diana Pottecher", "Angus Roberts", "Patrick Ruch", "Matthias Samwald", "Priscille Schneller", "Veronika Stefanov", "Miguel Angel Tinte", "Zdeňka Urešová", "Alejandro Vargas", "Dina Vishnyakova" ]
https://www.semanticscholar.org/paper/728fb39c077437138d46d3c43cbaa0022bb4d828/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
Improved Spelling Error Detection and Correction for Arabic
Vylepšená detekce a korekce pravopisných chyb v arabštině
A spelling error detection and correction application is based on three main components: a dictionary (or reference word list), an error model and a language model. While most of the attention in the literature has been directed to the language model, we show how improvements in any of the three components can lead to significant cumulative improvements in the overall performance of the system. We develop our dictionary of 9.3 million fully inflected Arabic words from a morphological transducer and a large corpus, cross validated and manually revised. We improve the error model by analysing error types and creating an edit distance re-ranker. We also improve the language model by analysing the level of noise in different sources of data and selecting the right subset to train the system on. Testing and evaluation experiments show that our system significantly outperforms Microsoft Word 2010, OpenOffice Ayaspell and Google Document.
Systémy pro detekci a korekci pravopisných chyb jsou obvykle založeny na třech komponentách: slovníku, chybovém modelu a jazykovém modelu. Zatímco většina ostatních prací se zaměřuje především na jazykový model, v této práci se ukazujeme, že vylepšení jakékoliv z těchto komponent může vést k významnému zlešení celého systému.
[ "Mohammed Attia", "Pavel Pecina", "Younes Samih", "Khaled Shaalan", "Josef van Genabith" ]
https://www.semanticscholar.org/paper/3186509b5f8f7e26bfe366c71c8438b048e19936/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
Rule-based Machine Translation between Indonesian and Malaysian
N/A
We describe the development of a bidirectional rule-based machine translation system between Indonesian and Malaysian (id-ms), two closely related Austronesian languages natively spoken by approximately 35 million people. The system is based on the re-use of free and publicly available resources, such as the Apertium machine translation platform and Wikipedia articles. We also present our approaches to overcome the data scarcity problems in both languages by exploiting the morphology similarities between the two.
(Not yet available. English version repeated) We describe the development of a bidirectional rule-based machine translation system between Indonesian and Malaysian (id-ms), two closely related Austronesian languages natively spoken by approximately 35 million people. The system is based on the re-use of free and publicly available resources, such as the Apertium machine translation platform and Wikipedia articles. We also present our approaches to overcome the data scarcity problems in both languages by exploiting the morphology similarities between the two.}
[ "Raymond Hendy Susanto", "Septina Dian Larasati", "Francis M. Tyers" ]
https://www.semanticscholar.org/paper/1c555fb678b42066494ada158b41a3a7b042e6a2/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
Korektor – A System for Contextual Spell-checking and Diacritics Completion
Korektor – systém pro kontextovou kontrolu pravopisu a doplňování diakritiky
We present Korektor – a flexible and powerful purely statistical text correction tool for Czech that goes beyond a traditional spell checker. We use a combination of several language models and an error model to offer the best ordering of correction proposals and also to find errors that cannot be detected by simple spell checkers, namely spelling errors that happen to be homographs of existing word forms. Our system works also without any adaptation as a diacritics generator with the best reported results for Czech text. The design of Korektor contains no language-specific parts other than trained statistical models, which makes it highly suitable to be trained for other languages with available resources. The evaluation demonstrates that the system is a state-of-the-art tool for Czech, both as a spell checker and as a diacritics generator. We also show that these functions combine into a potential aid in the error annotation of a learner corpus of Czech.
Představujeme Korektor – flexibilní statistický nástroj pro opravu českých textů, jehož schopnosti přesahují tradiční nástroje pro kontrolu pravopisu. Korektor využívá kombinace jazykových modelů a chybového modelu jak k tomu, aby setřídil pořadí nabízených náhrad pro neznámé slovo podle pravděpodobnosti výskytu na daném místě v textu, tak také, aby nalezl i překlepy, které se nahodile shodují s existujícím českým slovním tvarem. Prostou náhradou chybového modelu náš pracuje Korektor také jako systém pro doplnění diakritiky („oháčkování textu“) s nejvyšší publikovanou úspěšností. Systém neobsahuje žádné významné jazykově specifické komponenty s výjimkou natrénovaných statistických modelů. Je tedy možné jej snadno natrénovat i pro jiné jazyky. Ukážeme, jakých zlepšení náš systém dosahuje v porovnání se stávajícími českými korektory pravopisu i systémy pro doplnění diakritiky. Ukážeme také, že kombinace těchto schopností pomáhá při anotaci chyb v korpusu češtiny jako druhého jazyka.
[ "Michal Richter", "Pavel Straňák", "Alexandr Rosen" ]
https://www.semanticscholar.org/paper/06e75c3bf89fa3c6fc78f279e0ed0977b9e03b79/
null
null
null
null
null
null
null
null
null
null
null
null
cs
2012
Czech and Machine Translation
Čeština a strojový překlad
An up-to-date broad study of various approaches of machine translation and translation evaluation techniques, highlighting their advantages and disadvantages. The book is written in Czech and also primarily focuses on aspects of Czech, documenting that Czech is not only interestingly complex but it is also well supplied with linguistic tools and data.
Knížka přináší aktuální souborný popis problematiky strojového překladu, různých přístupů k němu a jejich výhod a nevýhod a zabírá se i otázkami vyhodnocování kvality překladu. Čeština jako podkladový jazyk potvrzuje svou vynikající roli: pro podchycení českých jazykových jevů je zapotřebí bohaté teorie, přitom je pro ale češtinu dostatek nástrojů a anotovaných i neanotovaných textových dat.
[ "Ondřej Bojar" ]
https://www.semanticscholar.org/paper/5468fa4da5e548fee0c30c28b60e297b5838f183/
null
null
null
null
null
null
null
null
null
null
null
null
cs
2012
The generation of machine translation
Generace strojového překladu
A talk on machine translation and a deliberation on the progress and regress it can lead to.
Přednáška o strojovém překladu spojená se zamyšlením, k jaké regeneraci či degeneraci díky překladu spějeme.
[ "Ondřej Bojar" ]
https://www.semanticscholar.org/paper/395de0bd3837fdf4b4b5e5f04835bcc69c279481/
null
null
null
null
null
null
null
null
null
null
null
null
cs
2012
A brief intro to machine translation
Jak se dělá strojový překlad
An introduction to the task of machine translation, current methods and their inherent limitations for those interested in computer science.
Seznámení zájemců o informatiku s úlohou strojového překladu, současnými metodami a jejich omezeními.
[ "Ondřej Bojar" ]
https://www.semanticscholar.org/paper/1aa94f75e9cfb7842e697a866f68971b058f9220/
null
null
null
null
null
null
null
null
null
null
null
null
cs
2012
Machine Translation
Strojový překlad
This is short article describing the field of machine translation to a broad audience.
Stručný článek představuje problematiku strojového překladu širšímu publiku.
[ "Ondřej Bojar" ]
https://www.semanticscholar.org/paper/1518039b5001f1836565215eb047526b3ac7f462/
null
null
null
null
null
null
null
null
null
null
null
null
cs
2012
Can machine translation do the homework for you?
Udělá za vás strojový překlad domácí úkol?
An introduction to machine translation for high school students and their teachers.
Seznámení studentů středních škol a jejich pedagogů s problematikou strojového překladu.
[ "Ondřej Bojar" ]
https://www.semanticscholar.org/paper/d48bc188dd1d5a23825eda1c360348f6e4f65ef4/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
Manually ranked outputs of Czech-Slovak translations
Manuálne ohodnotené česko-slovenské preklady
Data from three sources (part of Acquis, WMT test set and sentences selected from the set of books) translated by 5 machine translation systems (Česílko, Česílko 2, Google Translate and Moses with different settings) from Czech to Slovak and evaluated by three annotators. The translations were manually ordered according to their quality.
Dáta z troch zdrojov (časť korpusu Acquis, časť testovacej sady z WMT a vety vybrané z kníh) automaticky preložené pomocou piatich systémov (Česílko, Česílko 2, Google Translate a Moses s rôznymi nastaveniami) z češtiny do slovenčiny a ohodnotené troma anotátormi. Preklady boli ručne zoradené podľa ich kvality.
[ "Ondřej Bojar", "Petra Galuščáková" ]
https://www.semanticscholar.org/paper/f7819822454ac8ea58c29a4d6f69c75bedf70de2/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
Probes in a Taxonomy of Factored Phrase-Based Models
Sondy v taxonomii faktorových modelů frázového překladu
We introduce a taxonomy of factored phrase based translation scenarios and conduct a range of experiments in this taxonomy. We point out several common pitfalls when designing factored setups. The paper also describes our WMT12 submissions CU-BOJAR and CU-POOR-COMB.
Představujeme taxonomii pro faktorové modely frázového překladu a provádíme sérii experimentů s konfiguracemi z navržené taxonomie. Odhalujeme přitom řadu chyb v návrhu překladových modelů, jichž je vhodné se vyvarovat. Článek slouží rovněž jako popis našich systémů CU-BOJAR a CU-POOR-COMB v soutěži WMT12.
[ "Ondřej Bojar", "Bushra Jawaid", "Amir Kamran" ]
https://www.semanticscholar.org/paper/28c81a31868940fbc25777e4127d212acaa6be0e/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
Towards a Predicate-Argument Evaluation for MT
K predikátově-argumentovému vyhodnocování strojového překladu
HMEANT (Lo and Wu, 2011a) is a manual MT evaluation technique that focuses on predicate-argument structure of the sentence. We relate HMEANT to an established lin- guistic theory, highlighting the possibilities of reusing existing knowledge and resources for interpreting and automating HMEANT. We apply HMEANT to a new language, Czech in particular, by evaluating a set of English- to-Czech MT systems. HMEANT proves to correlate with manual rankings at the sentence level better than a range of automatic met- rics. However, the main contribution of this paper is the identification of several issues of HMEANT annotation and our proposal on how to resolve them.
HMEANT (Lo, Wu, 2011) je technika ručního hodnocení kvality strojového překladu založená na predikátově-argumentové struktuře věty. V článku dáváme HMEANT do souvislosti se zavedeným Funkčním generativním popisem a současně poprvé aplikujeme HMEANT na jiný jazyk než angličtinu, konkrétně na češtinu.
[ "Ondřej Bojar", "Dekai Wu" ]
https://www.semanticscholar.org/paper/90a509dc19771a38b12ac270a0c615ee38bf88bd/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
Scientific Report on Rich Tree-Based SMT
Technická zpráva o bohatém stromovém statistickém strojovém překladu
The final technical report on rich tree-based translation for the EuroMatrixPlus project.
Závěrečná technická zpráva grantu EuroMatrixPlus popisující strojový překlad prostřednictvím stromových struktur s bohatou anotací.
[ "Ondřej Bojar", "Mauro Cettolo", "Silvie Cinková", "Philipp Koehn", "Miroslav Týnovský", "Zdeněk Žabokrtský" ]
https://www.semanticscholar.org/paper/cc30577ccc88d4fe5989f4a2d2e5dd12315c15ef/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
The Joy of Parallelism with CzEng 1.0
Radost z paralelního korpusu CzEng 1.0
CzEng 1.0 is an updated release of our Czech-English parallel corpus, freely available for non-commercial research or educational purposes. In this release, we approximately doubled the corpus size, reaching 15 million sentence pairs (about 200 million tokens per language). More importantly, we carefully filtered the data to reduce the amount of non-matching sentence pairs. CzEng 1.0 is automatically aligned at the level of sentences as well as words. We provide not only the plain text representation, but also automatic morphological tags, surface syntactic as well as deep syntactic dependency parse trees and automatic co-reference links in both English and Czech. This paper describes key properties of the released resource including the distribution of text domains, the corpus data formats, and a toolkit to handle the provided rich annotation. We also summarize the procedure of the rich annotation (incl. co-reference resolution) and of the automatic filtering. Finally, we provide some suggestions on exploiting such an automatically annotated sentence-parallel corpus.
CzEng 1.0 je aktualizovaná verze česko-anglického paralelního korpusu, volně použitelného pro nekomerční použití. Oproti předchozí verzi je velikost korpusu dvojnásobně zvětšena na 15 milionů větných párů (řádově 200 milionů slov pro každý jazyk). Data jsou pečlivě profiltrována, aby se omezil výskyt neodpovídajících si větných párů apod. CzEng 1.0 je automaticky zarovnán po větách i po slovech. Krom čistě textové verze dáváme k dispozici anotaci korpusu na několika rovinách: morfologické, větně členské (analytické, povrchová závislostní syntax) a tektogramatické (hloubková syntax). Obsažena je také automatická anotace koreference pro oba jazyky.
[ "Ondřej Bojar", "Zdeněk Žabokrtský", "Ondřej Dušek", "Petra Galuščáková", "Martin Majliš", "David Mareček", "Jiří Maršík", "Michal Novák", "Martin Popel", "Aleš Tamchyna" ]
https://www.semanticscholar.org/paper/62563ffee9396a2601b2293690e6498545817210/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
Additional German-Czech reference translations of the WMT'11 test set
Dodatečné německo-české referenční překlady testovacích dat WMT'11
Additional three Czech reference translations of the whole WMT 2011 data set (http://www.statmt.org/wmt11/test.tgz), translated from the German originals. Original segmentation of the WMT 2011 data is preserved.
Tři přídavné české referenční překlady celé datové sady WMT 2011 (http://www.statmt.org/wmt11/test.tgz), přeložené z němčiny. Původní segmentace dat z WMT 2011 byla zachována.
[ "Ondřej Bojar", "Daniel Zeman", "Ondřej Dušek", "Jana Břečková", "Hana Farkačová", "Pavel Grošpic", "Kristýna Kačenová", "Eva Knechtlová", "Anna Koubová", "Jana Lukavská", "Petra Nováková", "Jana Petrdlíková" ]
https://www.semanticscholar.org/paper/9e0966e6d7e2120e2420f64f065611d5f48ff3aa/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
The Czech Language in the Digital Age
Čeština v digitálním věku
This white paper presents the state of language technology support for the Czech language. It is a part of a series that analyses the available language resources and technologies for 30 European languages.
Bílá kniha prezentuje stav podpory jazykových technologií pro češtinu. Je částí série, která analyzuje dostupné jayzkové zdroje pro 30 evropských jazyků.
[ "Ondřej Bojar", "Silvie Cinková", "Jan Hajič", "Barbora Hladká", "Vladislav Kuboň", "Jiří Mírovský", "Jarmila Panevová", "Nino Peterek", "Johanka Spoustová", "Zdeněk Žabokrtský" ]
https://www.semanticscholar.org/paper/ae72d7a1fca849d9cbce7e62982112d62043b6e1/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
eman
eman
Eman (“experimental manager”) is a software tool capable of maintaining large networks of mutually interconnected experiment processing large datasets, typically on a computational cluster. Eman was developed as an infrastructure for statistical machine translation but it can be used for other tasks as well.
Eman („experimentální manažer“) je softwarový nástroj, který umožňuje řídit rozsáhlé soubory vzájemně provázaných experimentů, při kterých se zpracovávají velké datové soubory, typicky na výpočetním clusteru. Byl vyvinut jako infrastruktura pro statistický strojový překlad, ale uplatní se i v jiných úlohách.
[ "Ondřej Bojar", "Aleš Tamchyna" ]
https://www.semanticscholar.org/paper/62dc469cd3993c8a2a6adeaeae9cc60c7544e4bc/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
Wild Experimenting in Machine Translation
Divoké experimentování se strojovým překladem
A tutorial on conducting a very broad range of experiments (for MT in particular).
Návod, jak provádět širokou sérii experimentů (se zaměřením na strojový překlad).
[ "Ondřej Bojar", "Aleš Tamchyna", "Jan Berka" ]
https://www.semanticscholar.org/paper/caf97def89a26b50588e10527b6f24a7cace3557/
null
null
null
null
null
null
null
null
null
null
null
null
cs
2012
Linguistics at the Faculty of Mathematics and Physics?
Lingvistika na Matematicko-fyzikální fakultě?
UFAL is a department well disposed to both linguistics and machine learning. It's well disposed to statistics as well because statistics is a key part of machine learning.
ÚFAL je pracovištěm, které je přátelsky nakloněné jak lingvistice, tak strojovému učení. Protože statistika je klíčovým pojmem strojového učení, je ÚFAL přátelsky spřízněn i se statistikou.
[ "Jan Hajič", "Barbora Vidová Hladká", "Jarmila Panevová" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
Announcing Prague Czech-English Dependency Treebank 2.0
Ohlašujeme Pražský česko-anglický závislostní korpus 2.0
We introduce a substantial update of the Prague Czech-English Dependency Treebank, a parallel corpus manually annotated at the deep syntactic layer of linguistic representation. The English part consists of the Wall Street Journal (WSJ) section of the Penn Treebank. The Czech part was translated from the English source sentence by sentence. This paper gives a high level overview of the underlying linguistic theory (the so-called tectogrammatical annotation) with some details of the most important features like valency annotation, ellipsis reconstruction or coreference.
Představujeme významnou aktualizaci Pražského česko-anglického závistostního korpusu (PCEDT), paralelního korpusu, který je ručně anotován na hloubkové syntaktické rovině.
[ "Jan Hajič", "Eva Hajičová", "Jarmila Panevová", "Petr Sgall", "Ondřej Bojar", "Silvie Cinková", "Eva Fučíková", "Marie Mikulová", "Petr Pajas", "Jan Popelka", "Jiří Semecký", "Jana Šindlerová", "Jan Štěpánek", "Josef Toman", "Zdeňka Urešová", "Zdeněk Žabokrtský" ]
https://www.semanticscholar.org/paper/334db7cbffc537bb4d9b25e0dd128e6434691af6/
null
null
null
null
null
null
null
null
null
null
null
null
cs
2012
Coordination vs. determination (Form or meaning?)
Koordinace vs. determinace (Forma nebo význam?)
The question about boundaries of so-called "hypotactical coordination" is discussed in connection with three types of constructions: (i) asociative constructions vs simple conjunctions, (ii) so-called "false" subordinated clauses, (iii) constructions introduced by the expressions "místo" (instead), "kromě" (beside/with exception).
V příspěvku se klade otázka, zda jsme oprávněni u typů (i) - (iii) mluvit o asymetrickém vztahu "hypotaktické koordinace". Jde o typy (i) asociativ vs slučovací koordinace, (ii) tzv. nepravé věty vedlejší, (iii) konstrukce uvozené výrazy "místo, kromě, mimo". Dochází se k názoru, že na hypotaktickou koordinaci kandiduje část příkladů typu (a).
[ "Jarmila Panevová" ]
https://www.semanticscholar.org/paper/b5ebc1bce60720049719511fdc012b9a3ab7f283/
null
null
null
null
null
null
null
null
null
null
null
null
cs
2012
The noticable loss in world Slavistics
Světová slavistika utrpěla citelné ztráty
These obituaries are devoted to couple of famouse linguists, first of all Slavists: prof. Rudolf Růžička from the University of Leipzig (1920-2011) and to prof. Milka Ivic, professor of the University in Novi Sad and Institute of Serbian Language in Beograd (1923-2011).
Jde o nekrology věnované prof. Rudolfu Růžičkovi z Univerzity v Lipsku, předním představiteli teoretické a formální lingvistiky v Německu (1920-2011) a prof. Milce Ivicové z Univerzity v Nověm Sadu a z Institutu srbského jazky SAVU v Bělehradě (1923-2011).
[ "Jarmila Panevová" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
ru
2012
On Assymetry of the Deep and Surface Representation of the Sentence
Asymetrie mezi hloubkovou a povrchovou reprezentací věty (na příkladu dvou typů českých příslovečných určení)
The Czech constructions of comparison and the constructions introduced by kromě/mimo are analyzed. For their deep structure it is necessary to expand them into the embedded predication, which is considered to be a grammaticalized ellipsis. The two new syntactic relations are introduced (ADDIT - an addition, EXC - an exception). They are connected with the ambiguity of kromě constructions.
Ve stati se analyzují české konstrukce srovnávací a konstrukce uvedené výrazy kromě/mimo. Navrhuje se jejich hloubková (tektogramatická) reprezentace rozvinutá v zapuštěnou predikaci a označovaná jako gramatikalizovaná elipsa. Zavádí se dva nové syntaktické vztahy (adice - ADDIT a výjimka - EXC) související s víceznačností konstrukcí s kromě.
[ "Jarmila Panevová", "Marie Mikulová" ]
null
null
null
Ассиметрии между глубинным и поверхностным преставлением предложения (на примере двух типов обстоятельств в чешском языке)
В статье анализируются чешские обстоятельства сравнения и обстоятельства введенные выражениями kromě (mimo). Предлагается их глубинное представление в виде вставленной предикации, считаемой грамматикализованным эллипсисом. Вводятся два новыцх типа синтактических связей (ADDIT и EXC).
null
null
null
null
null
null
null
null
cs
2012
Two views on the Development of the Czech Post-War Syntactic Research
Dva pohledy na vývoj českého poválečného syntaktického myšlení
The metodological approaches used in Czech syntactic research since the 40-ies of the last century are analyzed. The Daneš´s Two-Level Valency Syntax and Sgall´s Functional Generative Description are evaluated as most comprehensive models in the Czech syntax.
V článku se hodnotí metodologické směry užívané v české syntaxi od 40. let minulého století. Jako nejvlivnější modely se hodnotí Danešova Dvourovinná valenční syntax a Sgallův Funkční generativní popis.
[ "Jarmila Panevová", "Petr Karlík" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
Syntactic annotation of transcriptions in the Czech Academic Corpus: Then and now
Syntaktická anotace přepisů mluvené řeči z Českého akademického korpusu: tehdy a nyní
Corpus annotation plays an important role in linguistic analysis and computa-tional processing of both written and spoken language. Syntactic annotation of spoken texts becomes clearly a topic of considerable interest nowadays, driven by the desire to improve auto-matic speech recognition systems by incorporating syntax in the language models, or to build language under-standing applications.
Korpusová anotace je důležitou součástí lingvistické analýzy a počítačového zpracování jazyka. Tento článek se zabývá problémy spojenými se syntaktickou anotací mluvených textů na pozadí syntaktické anotace ČAKu.
[ "Barbora Vidová Hladká", "Zdeňka Urešová" ]
https://www.semanticscholar.org/paper/cb6ce358b0b4ce3deb4a75c5213af9bd2d6e6505/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
Conversion of Copenhagen Dependency Treebank into Treex
Převod Kodaňského závislostního korpusu do systému Treexu
I would like to present my recent efforts on converting the Copenhagen Dependency Treebank (CDT) data into Treex. CDT is a multilingual treebank developed at CBS. Treex is a multi-purpose Natural Language Processing software framework developed at Charles University in Prague. Treex is used in a number of projects aimed at building language data resources as well as at developing NLP applications such as Machine Translation systems. The talk will have two parts. In the first part I will give an overview of the conversion procedure, with focus on pecularities of CDT that made the conversion a challenging task. In the second part I will show how the CDT data in the new format can be browsed and further processed.
V přednášce představím software, který jsem vyvinul pro převod Kodaňského závislostního korpusu (CDT) do systému Treex. CDT je vícejazyčný korpus vyvinutý v Copenhagen Business School. Treex je multilinguální softwarový systém vyvinutý na Karlově univerzitě v Praze. Treex se používá mimo jiné pro vývoj systémů strojového překladu. Přednáška bude mít dvě části. V první představím konverzní proceduru. Ve druhé části předvedu možné způsoby, jak s daty v novém formátu pracovat.
[ "Zdeněk Žabokrtský" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
Machine Translation using Dependency Trees
Strojový překlad s využitím závislostních stromů
Předložená práce je zaměřená na využití stromových syntaktických struktur jakožto reprezentace věty přirozeného jazyka v experimentálním systému pro anglicko-český strojový překlad.
The present work focuses on using tree-shaped syntactic structures as an intermediate sentence representation in an experimental English-Czech machine translation system.
[ "Zdeněk Žabokrtský" ]
https://www.semanticscholar.org/paper/7fcccd2c11da75f2ae4cfc2b7be197423fc22d0e/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
Tailored Feature Extraction for Lexical Disambiguation of English Verbs Based on Corpus Pattern Analysis
Rysy šité na míru anglickým sloves°um pro automatickou lexikální disambiguaci pomocí Corpus Pattern Analysis
We give a report on a detailed study of automatic lexical disambiguation of 30 sample English verbs. We have trained and evaluate several statistical classifiers that use both morphosyntactic and semantic features to assign semantic patterns according to a pattern lexicon. Our system of semantic classification draws on the Corpus Pattern Analysis (CPA) — a novel lexicographic method that seeks to cluster verb uses according to the morpho-syntactic, lexical and semantic/pragmatic similarity of their contexts rather than their grouping according to abstract semantic definitions. In this paper we mainly concentrate on the procedures for feature extraction and feature selection. We show that features tailored to particular verbs using contextual clues given by the CPA method and explicitly described in the pattern lexicon have potential to significantly improve accuracy of supervised statistical classifiers.
Předkládáme detailní studii automatické lexikální disambiguace na pilotním vzorku třiceti anglických sloves za použití lexikonu vzorů slovesných užití (patterns), který vychází z Corpus Pattern Analysis (CPA). Tato inovátorská lexikografická metoda namísto na abstraktních definicích jednotlivých významů staví na souhře morfosyntaktické, lexikální a sémantické/pragmatické podobnosti slovesných užití. Natrénovali jsme několik statistických klasifikátorů na rozpoznávání těchto vzorů. Klasifikátory využívají jak morfosyntaktických, tak sémantických rysů. V naší studii se soustředíme na procedury pro extrakci rysů, jejich výběr a jejich evaluaci. Ukazujeme, že rysy na míru uzpůsobené jednotlivým slovesům, jež jsou implicitně obsaženy v definici každého vzoru v lexikonu, mají potenciál významně zvýšit přesnost statistických klasifikátorů s učitelem.
[ "Martin Holub", "Vincent Kríž", "Silvie Cinková", "Eckhard Bick" ]
https://www.semanticscholar.org/paper/801ce644f9a2883d774f1761b437e765da09da7e/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
On scalarity in information structure
Škálovitost v aktuálním členění
Analysis of different approaches to the concept and the description of hierarchy of element from the point of view of topic-focus articulation.
Rozbor různých přístupů k pojetí a popisu stupňovitosti z hlediska aktuálního členění.
[ "Eva Hajičová" ]
https://www.semanticscholar.org/paper/3f9a6e16010d78997699bbb9eb51e90fd005c4c4/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
The Functional Generative Description as a functionally motivated formal model of language
Funkční generativní popis jako funkčně motivovaný formalní model jazyka
The article is about main principles of the explicit description of language based on dependency syntax.
Hlavní principy explicitního pražského závislostního modelu jazyka postupující od funkce k formě.
[ "Eva Hajičová" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
Topic-Focus revisited (Through the eyes of the Prague Dependency Treebank)
Znovu k aktuálnímu členění (Očima Pražského závislostního korpusu)?
A comparison of two theoretical approaches to the information structure and the application of the Praguian approach in an annotated corpus of Czech.
Srovnání dvou teoretických přístupů k aktuálnímu členění a uplatnění pražského přistupu v anotovaném počítačovém korpusu češtiny.
[ "Eva Hajičová" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
Vilém Mathesius and Functional Sentence Perspective, and beyond
Vilém Mathesius a aktuální členění, a související otázky
A comparison of different Praguian approaches to information structure, from Vilem Mathesius up today.
Srovnání přístupu k aktuálnímu členění v české lingvistice, od základní práce Mathesiovy až po práce současné.
[ "Eva Hajičová" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
What we have learned from complex annotation of topic-focus articulation in a large Czech corpus
Co jsme se naučili z anotace aktuálního členění ve velkém českém korpusu.
Linguistic issues that had to be analyzed and resolved during the annotation of information structure in the Prague Dependency Treebank.
Článek shrnuje lingvistické problémy, na které jsme narazili a které řešili anotátoři při anotování informační struktury ve větách Pražského závislostního korpusu.
[ "Eva Hajičová" ]
https://www.semanticscholar.org/paper/32e89d0406017b305d92fa7755965fc78ee9afcb/
null
null
null
null
null
null
null
null
null
null
null
null
cs
2012
From the roots to the trees: A story of annotated corpora of Czech
Od kořenů ke stromům: příběh anotovaných korpusů češtiny
Academic corpus of the Czech language created in the Institute of Czech Language belongs to the first annotated corpora of natural languages ever in existence. This activity was one of the stimuli for the development of the richly annotated corpus of the Prague Dependency Treebank.
Akademický korpus češtiny zpracovaný v ÚJČ AV byl jedním z prvních anotovaných počítačových korpusů přirozeného jazyka, a to na rovině morfosyntaktické. Odtud vedla cesta k bohatě anotovanému Pražskému závislostnímu korpusu.
[ "Eva Hajičová", "Barbora Hladká" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
Dspace Modifications for Clarin Compatible (EPIC) Handle Service
Úpravy Dspace pro služby Handle kompatibilní s Clarinem (EPIC)
This software is a modification and configuration of Dspace 1.6 that allows Clarin centres that employ Dspace as their repository software to use Clarin recommended Handle Service from EPIC Consortium within Dspace.
Tento software je modifikací a konfigurací Dspace 1.6, která umožňuje centrům Clarin, která používají Dspace jako svůj software pro repozitáře, využít službu Handle od konzorcia EPIC, doporučenou Clarinem, v rámci Dspace.
[ "Petr Pajas" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
Domain Adaptation of Statistical Machine Translation using Web-Crawled Resources: a Case Study
Doménová adaptace statistického strojového překladu s pomocí dat získaných z webu: případová studie
We tackle the problem of domain adaptation of Statistical Machine Translation by exploiting domain-specific data acquired by domain-focused web-crawling. We design and evaluate a procedure for automatic acquisition of monolingual and parallel data and their exploitation for training, tuning, and testing in a phrase-based Statistical Machine Translation system. We present a strategy for using such resources depending on their availability and quantity supported by results of a large-scale evaluation on the domains of Natural Environment and Labour Legislation and two language pairs: English--French, English--Greek. The average observed increase of BLEU is substantial at 49.5% relative.
V této práci se zabýváme doménovou adaptací statistického strojového překladu s využitím dat automaticky stažených z internetu. Experimenty jsou provedeny na doménách životního prostředí a pracovní legislativy.
[ "Pavel Pecina", "Antonio Toral", "Vassilis Papavassiliou", "Prokopis Prokopidis", "Josef Genabith" ]
https://www.semanticscholar.org/paper/7e842596f9bec3b6f9d7a1ebff479e72b663e775/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
Simple and Effective Parameter Tuning for Domain Adaptation of Statistical Machine Translation
Jednoduchá a efektivní optimalizace parametrů pro doménovou adaptaci statistického strojového překladu
Current state-of-the-art Statistical Machine Translation systems are based on log-linear models that combine a set of feature functions to score translation hypotheses during decoding. The models are parametrized by a vector of weights usually optimized on a set of sentences and their reference translations, called development data. In this paper, we explore a (common and industry relevant) scenario where a system trained and tuned on general domain data needs to be adapted to a specific domain for which no or only very limited in-domain bilingual data is available. It turns out that such systems can be adapted successfully by re-tuning model parameters using surprisingly small amounts of parallel in-domain data, by cross-tuning or no tuning at all. We show in detail how and why this is effective, compare the approaches and effort involved. We also study the effect of system hyperparameters (such as maximum phrase length and development data size) and their optimal values in this scenario.
Současné systémy statistického strojového překladu jsou založeny na logaritmicko-lineárních modelech, které pro hodnocení překladových hypotéz ve fázi dekódování kombinují sadu příznakových funkcí. Tyto modely jsou parametrizovány vektorem vah, které se optimalizují na tzv. vývojových datech, tj. množině vět a jejich referenčních překladů. V tomto článku se zabýváme (častou a pro průmyslové nasazení relevantní) situací, kdy je třeba překladový systém natrénovaný na datech z obecné domény adaptovat na nějakou specifickou doménu, pro kterou jsou k dispozici paralelní data jen ve velice omezeném (či žádném) množství. Ukazujeme, že takové systémy mohou být vhodně adaptovány pomocí optimalizace parametrů za použití jen překvapivě malého množství paralelních doménově-specifických dat nebo tzv. křížovou optimalizací. Možností je také nepoužití optimalizace vůbec. Jednotlivé přístupy analyzujeme a porovnáváme jejich cekovou náročnost. Dále se zabýváme analýzou systémových hyperparametrů (např. maximální délkou frází a velikostí vývojových dat) a jejich optimalizací.
[ "Pavel Pecina", "Antonio Toral", "Josef van Genabith" ]
https://www.semanticscholar.org/paper/af6d744b606b7cee709e0999cd3d1c0045ddb1c5/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
Building the PDT-VALLEX valency lexicon
Vznik valenčního slovníku PDT-Vallex
In our contribution, we relate the development of a richly annotated corpus and a computational valency dictionary. Our valency dictionary has been created as a “byproduct” of the annotation of the Prague Dependency Treebank (PDT) but it became an important resource for further linguistic research as well as for computational processing of the Czech language.
Tento příspěvek se zabývá vztahem korpusu a valenčního slovníku. Slovník vznikl jako vedlejší produkt anotace Pražského závislostního korpusu, stal se důležitým zdrojem jak pro další lingvistický výzkum, tak pro počítačové zpracování češtiny.
[ "Zdeňka Urešová" ]
https://www.semanticscholar.org/paper/edacf3935a6ad4a8631be2de920f1d3e8d80f0cb/
null
null
null
null
null
null
null
null
null
null
null
null
cs
2012
Valency of nouns derived from verbs with an object expressed by Genitive
Valence dějových substantiv odvozených od sloves s předmětovým genitivem
In the present paper, adnominal counterparts of adverbal objects expressed by prepositionless Genitive (Gen; e.g. obávat se čeho ‘to-be-afraid of-sth’, dotázat se koho ‘to-ask of-sb’) are studied. The intention was to find out if Czech nouns derived from verbs by non-productive means can be modified by Patient (PAT) or Addressee (ADDR) expressed by Gen as well; neither other papers nor valency dictionaries mention this form of the participants. It has turned out that several nouns modified by PAT expressed by Gen can rarely be found in the corpus (e.g. CNC SYN2005, cf. odvaha spolupráce ‘courage of-cooperation’, jeho dotek puku ‘his touch of-the-puck’; other forms of PAT, i.e. prepositional groups, sometimes also infinitive or an embedded objective clause, are more frequent). Factors that influence possibility or impossibility to be modified by PAT or ADDR expressed by Gen are typically connected with the type of the semantic group the noun belongs to (cf. the difference between nouns denoting “positive” vs. “negative” mental state or dispositions, e.g. naděje úspěchu ‘hope of-success’ vs. *obava následků ‘fear of-consequences’, or the tendency to avoid syntactic homonymy of Actor (ACT) and ADDR expressed by Gen, which is typical of nouns of saying, cf. dotaz kamaráda.ACT vs. *dotaz kamaráda.ADDR ‘question of-the-friend’).
Příspěvek se věnuje valenčním vlastnostem českých neproduktivně tvořených (dějových) substantiv odvozených od sloves s předmětovým genitivem. Prvotní motivací bylo zjistit, zda si tato substantiva (např. obava, dotaz, dotek) uchovávají původní adverbální genitivní vazbu, tj. GenAdnom (← GenAdverb). Valenční slovníky ani odborná literatura tuto vazbu u daných substantiv neuvádějí. V subkorpusech ČNK se u některých substantiv vyskytuje, např. naděje úspěchu, jeho dotek puku; vedle ostatních forem daného participantu (např. předložkových skupin) se jedná spíše o okrajovou možnost vyjádření. Za možností / nemožností užití GenAdnom (← GenAdverb) u dějových substantiv stojí různorodé faktory, související patrně především s příslušností daných substantiv k určité sémantické skupině, příp. podskupině (srov. pozitivní vs. negativní duševní stavy, např. odvaha spolupráce vs. *obava zkoušky, snaha vyhnout se strukturní homonymii u substantiv mluvení, např. *dotaz kamaráda.ADDR vs. dotaz kamaráda.ACT).
[ "Veronika Kolářová" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
cs
2012
Dialogy.Org 1.1
Dialogy.Org 1.1
Dialogy.Org system is a software tool designed primarily to support the creation and sharing of annotated audio-visual data by linguistic community. The software tool allows editing and searching in the transcripts of audio-visual recordings of dialogues. The dynamic web application provides access for registered users to the digitised archive. Playing and exploring of selected parts is possible in the web browser.
Systém Dialogy.Org je softwarový nástroj určený především k podpoře vytváření a sdílení anotovaných audio-video dat lingvistickou komunitou. Dialogy.Org umožňuje formou webového rozhraní editaci a vyhledávání v přepisech audio-visuálních nahrávek dialogů. Vyhledané úseky textu je možné přehrávat a analyzovat pomocí webového prohlížeče.
[ "Nino Peterek" ]
https://www.semanticscholar.org/paper/7d08e0e5dcaf12ee56a4bc4acae9b273b3fda5ef/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
Does Tectogrammatics Help the Annotation of Discourse?
Pomáhá tektogramatika při anotaci diskurzních vztahů?
In the following paper, we discuss and evaluate the benefits that deep syntactic trees (tectogrammatics) and all the rich annotation of the Prague Dependency Treebank bring to the process of annotating the discourse structure, i.e. discourse relations, connectives and their arguments. The decision to annotate discourse structure directly on the trees contrasts with the majority of similarly aimed projects, usually based on the annotation of linear texts. Our basic assumption is that some syntactic features of a sentence analysis correspond to certain discourselevel features. Hence, we use some properties of the dependency-based large-scale treebank of Czech to help establish an independent annotation layer of discourse. The question that we answer in the paper is how much did we gain by employing this approach.
V tomto příspěvku hodnotíme přínos, který představují syntacticko-sémantické stromy (tektogramatická rovina anotace) a celá bohatá anotace Pražského závislostního korpusu pro anotaci diskurzní struktury textu, tedy pro anotaci diskurzních vztahů, jejich konektorů a argumentů. Rozhodnutím anotovat diskurzní strukturu přímo na stromech se náš přístup liší od většiny podobně zaměřených projektů, které jsou obvykle založeny na anotaci lineárního textu. Naším základním předpokladem je, že některé syntaktické rysy větné analýzy odpovídají jistým rysům z roviny diskurzní struktury. Proto využíváme některé vlastnosti rozsáhlého závislostního korpusu češtiny k ustanovení nezávislé diskurzní anotační vrstvy. V tomto příspěvku odpovídáme na otázku, jaké výhody tento přístup přináší.
[ "Jiří Mírovský", "Pavlína Jínová", "Lucie Poláková" ]
https://www.semanticscholar.org/paper/1053a9f6e428ffe745004c8fed1b6d1d645946ce/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
CUNI: Feature Selection and Error Analysis of a Transition-Based Parser
CUNI: Výběr rysů a analýza chyb parseru založeného na přechodech
The paper describes our system used for dependency parsing of Hindi data during the shared task at Coling 2012.
Příspěvek popisuje systém použitý pro závislostní syntaktickou analýzu hindských dat v rámci soutěže na Colingu 2012.
[ "Daniel Zeman" ]
https://www.semanticscholar.org/paper/b1f53c39835ce4084980d832cbe2da2ff9e4ceff/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
Data Issues of the Multilingual Translation Matrix
Otázka dat ve vícejazyčné překladové matici
We describe our experiments with phrase-based machine translation for the WMT 2012 Shared Task. We trained one system for 14 translation directions between English or Czech on one side and English, Czech, German, Spanish or French on the other side. We describe a set of results with different training data sizes and subsets.
Popisujeme naše pokusy s frázovým strojovým překladem pro soutěž WMT 2012. Natrénovali jsme jeden systém pro 14 překladových párů mezi angličtinou nebo češtinou na jedné straně a angličtinou, češtinou, němčinou, španělštinou nebo francouzštinou na druhé straně. Popisujeme sadu výsledků s různými velikostmi trénovacích dat a jejich podmnožin.
[ "Daniel Zeman" ]
https://www.semanticscholar.org/paper/fec523feacbf6f8cdafeb1f7ac7d974b91768e46/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
Multilingual SMT: Data, Data, Data…
Vícejazyčný statistický strojový překlad: data, data, data…
Observations from this year's experiments at ÚFAL for the WMT shared translation task.
Postřehy z letošních pokusů na ÚFALu v rámci mezinárodní soutěže ve strojovém překladu.
[ "Daniel Zeman" ]
https://www.semanticscholar.org/paper/28ff558866d65449a8cc68373562a861bcbe505a/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
HamleDT: To Parse or Not to Parse?
HamleDT: Rozebrat, či nerozebrat?
We propose HamleDT – HArmonized Multi-LanguagE Dependency Treebank. HamleDT is a compilation of existing dependency treebanks (or dependency conversions of other treebanks), transformed so that they all conform to the same annotation style. While the license terms prevent us from directly redistributing the corpora, most of them are easily acquirable for research purposes. What we provide instead is the software that normalizes tree structures in the data obtained by the user from their original providers.
Představujeme HamleDT – harmonizovaný mnohojazyčný závislostní korpus. HamleDT je kolekce existujících závislostních korpusů (nebo do závislostí převedených jiných syntakticky anotovaných korpusů), transformovaných tak, aby odpovídaly jednotnému anotačnímu stylu. Licenční podmínky nám nedovolují dále šířit samotné korpusy, většina z nich je ovšem pro vědecké účely snadno dostupná a my nabízíme software, který tato data převede do normalizované podoby.
[ "Daniel Zeman", "David Mareček", "Martin Popel", "Loganathan Ramasamy", "Jan Štěpánek", "Zdeněk Žabokrtský", "Jan Hajič" ]
https://www.semanticscholar.org/paper/f07ad1ff6db0d43a6ab0be817bd5ee5daf852ba7/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
Intersecting Parallel Corpora
Výpočet průniku paralelních korpusů
The organizers of the annual Workshop on Machine Translation (WMT) prepare and distribute parallel corpora that can be used to train systems for the shared tasks. Two core types of corpora are the News Commentary corpus and the Europarl corpus. Both are available in several language pairs, always between English and another European language: cs-en, de-en, es-en and fr-en. The corpora are not multi-parallel. They come from the same source and there is significant overlap but still some sentences are translated to only a subset of the languages. The bi-parallel subsets do not all have the same number of sentence pairs. Such corpora cannot be directly used to train a system for e.g. de-cs (German-Czech). However, we can use English as a pivot language. If we identify the intersection of the English parts of cs-en and de-en, we can take the non-English counterparts of the overlapping English sentences to create a de-cs parallel corpus. That is what this software does.
Organizátoři každoročního Semináře o strojovém překladu (WMT) připravují a šíří paralelní korpusy, které lze použít při trénování systémů pro soutěžní úlohy. Mezi hlavní typy korpusů patří korpusy News Commentary a Europarl. Oba jsou k dispozici v několika jazykových párech, vždy mezi angličtinou a dalším evropským jazykem: cs-en, de-en, es-en a fr-en. Tyto korpusy nejsou paralelní přes více než dva jazyky. Pocházejí ze stejného zdroje a významně se překrývají, přesto však jsou některé věty přeloženy jen do některých jazyků. Dvojjazyčné paralelní podmnožiny nemají všechny stejný počet párů vět. Takové korpusy nemůžeme přímo nasadit při trénování systému pro např. de-cs (němčina-čeština). Můžeme nicméně použít angličtinu jako pivotní jazyk. Pokud rozpoznáme průnik anglických částí cs-en a de-en, můžeme použít jejich neanglické protějšky a vytvořit z nich paralelní korpus de-cs. Tuto úlohu plní tento software.
[ "Daniel Zeman", "Ondřej Bojar" ]
https://www.semanticscholar.org/paper/895551d3c850d427c4d8e69f3bb907305c81d507/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
VALLEX 2.6
VALLEX 2.6
The Valency Lexicon of Czech Verbs, Version 2.6 (VALLEX 2.6), is a collection of linguistically annotated data and documentation, resulting from an attempt at formal description of valency frames of Czech verbs. VALLEX 2.6 has been developed at the Institute of Formal and Applied Linguistics, Faculty of Mathematics and Physics, Charles University, Prague. VALLEX 2.6 provides information on the valency structure (combinatorial potential) of verbs in their particular senses. VALLEX 2.6 is a successor of VALLEX 1.0, extended in both theoretical and quantitative aspects (including corpus evidence).
Valenční slovník českých sloves VALLEX, verze 2.6 obsahující přibližně 2730 záznamů o lexémech pokrývajících cca. 6460 lexikálních jednotek (významů). Slovník je volně k dispozici pro účely výzkumu. Nová verze slovníku je - oproti starší verzi - kvalitativně i kvantitativně obohacena (zejm. jsou pro vybraná slovesa doplněny korpusové příkladů).
[ "Markéta Lopatková", "Václava Kettnerová", "Eduard Bejček", "Karolína Skwarska", "Zdeněk Žabokrtský" ]
https://www.semanticscholar.org/paper/5e707fe81dd223275f66f8771251f6a7ac2213bd/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
Annotation of sentence structure: Capturing the relationship between clauses in Czech sentences
Anotace větné struktury: Vztahy mezi klauzemi v českém souvětí
The focus of this article is on the creation of a collection of sentences manually annotated with respect to their sentence structure. We show that the concept of linear segments—linguistically motivated units, which may be easily detected automatically—serves as a good basis for the identification of clauses in Czech. The segment annotation captures such relationships as subordination, coordination, apposition and parenthesis; based on segmentation charts, individual clauses forming a complex sentence are identified. The annotation of a sentence structure enriches a dependency-based framework with explicit syntactic information on relations among complex units like clauses. We have gathered a collection of 3,444 sentences from the Prague Dependency Treebank, which were annotated with respect to their sentence structure (these sentences comprise 10,746 segments forming 6,341 clauses). The main purpose of the project is to gain a development data—promising results for Czech NLP tools (as a dependency parser or a machine translation system for related languages) that adopt an idea of clause segmentation have been already reported. The collection of sentences with annotated sentence structure provides the possibility of further improvement of such tools.
Článek se soustřeďuje na popis kolekce vět, u kterých byla ručně analyzována větná struktura. Ukazuje, že koncept založený na lineárních segmentech, které lze snadno automaticky detekovat, slouží jako dobrý základ pro identifikaci klauzí v češtině. Anotace segmentů zahrnuje takové jevy jako je závislost kauzí, jejich koordinace, apozice či parenteze. Anotace větné struktury doplňuje závislostní přístup k popisu jazyka o explicitní informaci o vztazích mezi klauzemi.
[ "Markéta Lopatková", "Petr Homola", "Natalia Klyueva" ]
https://www.semanticscholar.org/paper/be3397737b632ea6acbaf6201e10048bd17ed8ed/
null
null
null
null
null
null
null
null
null
null
null
null
cs
2012
We teach machines to speak Czech
Učíme stroje česky
Presentations to high school students within the "Day of professions". The aim was to introduce students to mathematical linguistics as an interesting field, and generally promote the study of natural sciences.
Popularizační prezentace pro studenty gymnázia Na Zatlance v rámci Dne profesí. Cílem bylo přiblížit studentům matematickou lingvistiku jako zajímavý obor a obecně propagovat studium přírodních věd.
[ "Jaroslava Hlaváčová" ]
https://www.semanticscholar.org/paper/65695d21c2eb1176617b7d649747c289729893b4/
null
null
null
null
null
null
null
null
null
null
null
null
cs
2012
Example of regular patterns of word formation in the automatic processing of Czech and Russian
Příklad pravidelných slovotvorných vzorců v automatickém zpracování češtiny a ruštiny
There is a set of prefixes in Czech as well as in Russian, which, added to imperfective verbs together with a reflexive pronoun, change the meaning of the verb in the same manner. This feature is so regular that could help automatically recognize words without using morphological dictionaries.
V češtině i v ruštině existuje množina předpon, jejichž připojením k nedokonavému slovesu a přidáním zvratného zájmena pozměníme význam původního slovesa vždy téměř stejným způsobem. Toho lze využít při automatickém rozpoznávání slovních tvarů, aniž by bylo třeba je ukládat do morfologických slovníků.
[ "Jaroslava Hlaváčová", "Anja Nedolužko" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
Maintaining consistency of monolingual verb entries with interannotator agreement
Udržování konzistence jednojazyčných slovesných hesel pomocí mezianotátorské shody
There is no objectively correct way to create a monolingual entry of a polysemous verb. By structuring a verb into readings, we impose our conception onto lexicon users, no matter how big a corpus we use in support. How do we make sure that our structuring is intelligible for others? We are performing an experiment with the validation of the fully corpus-based Pattern Dictionary of English Verbs (Hanks & Pustejovsky, 2005), created according to the lexical theory Corpus Pattern Analysis (CPA). The lexicon is interlinked with a large corpus, in which several hundred randomly selected concordances of each processed verb are manually annotated with numbers of their corresponding lexicon readings (“patterns”). It would be interesting to prove (or falsify) the leading assumption of CPA that, given the patterns are based on a large corpus, individual introspection has been minimized and most people can agree on this particular semantic structuring. We have encoded the guidelines for assigning concordances to patterns and hired annotators to annotate random samples of verbs cotained in the lexicon. Apart from measuring the interannotator agreement, we analyze and adjudicate the disagreements. The outcome is offered to the lexicographer as feedback. The lexicographer revises his entries and the agreement can be measured againg on a different random sample to test whether or not the revision has brought an improvement of the interannotator agreement score. A high interannotator agreement suggests that lexicon users are likely to find a pattern corresponding to a random verb use of which they seek explanation. A low agreement score gives a warning that there are patterns missing or vague. We focus on machine-learning applications, but we believe that this procedure is of interest even for quality management in human lexicography.
Článek se věnuje problematice sémantické analýzy sloves a lexikálnímu popisu sloves se zřetelem k vytvoření trénovacích dat pro automatický sémantický analyzátor. Vychází z metody Corpus Pattern Analysis a zkoumá její uplatnění v lexikálním popisu sloves pro účely NLP pomocí měření mezianotátorské shody.
[ "Silvie Cinková", "Lenka Smejkalová", "Anna Vernerová", "Jonáš Thál", "Martin Holub" ]
https://www.semanticscholar.org/paper/da38f978047af3c92c8571092238cae361ee5d5e/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
A database of semantic clusters of verb usages
Databáze sémantických shluků slovesných kontextů
We are presenting VPS-30-En, a small lexical resource that contains the following 30 English verbs: access, ally, arrive, breathe, claim, cool, crush, cry, deny, enlarge, enlist, forge, furnish, hail, halt, part, plough, plug, pour, say, smash, smell, steer, submit, swell, tell, throw, trouble, wake and yield. We have created and have been using VPS-30-En to explore the interannotator agreement potential of the Corpus Pattern Analysis. VPS-30-En is a small snapshot of the Pattern Dictionary of English Verbs (Hanks and Pustejovsky, 2005), which we revised (both the entries and the annotated concordances) and enhanced with additional annotations. It is freely available at http://ufal.mff.cuni.cz/spr. In this paper, we compare the annotation scheme of VPS-30-En with the original PDEV. We also describe the adjustments we have made and their motivation, as well as the most pervasive causes of interannotator disagreements.
Představujeme VPS-30-En, malý lexikální zdroj, který obsahuje následujících 30 sloves: access, ally, arrive, breathe, claim, cool, crush, cry, deny, enlarge, enlist, forge, furnish, hail, halt, part, plough, plug, pour, say, smash, smell, steer, submit, swell, tell, throw, trouble, wake and yield. VPs-30-En jsme vytvořili a používáme jej k výzkumu potenciálu mezianotátorské shody u sémantického značkování podle anotačního schématu CPA.
[ "Silvie Cinková", "Martin Holub", "Adam Rambousek", "Lenka Smejkalová" ]
https://www.semanticscholar.org/paper/db9379dc35d39d033902641ae6106bc49ce397a6/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
Managing Uncertainty in Semantic Tagging
Kontrola nejistoty v sémantickém značkování
Low interannotator agreement (IAA) is a well-known issue in manual semantic tagging (sense tagging). IAA correlates with the granularity of word senses and they both correlate with the amount of information they give as well as with its reliability. We compare different approaches to semantic tagging in WordNet, FrameNet, Prop- Bank and OntoNotes with a small tagged data sample based on the Corpus Pattern Analysis to present the reliable information gain (RG), a measure used to optimize the semantic granularity of a sense inventory with respect to its reliability indicated by the IAA in the given data set. RG can also be used as feedback for lexicographers, and as a supporting component of automatic semantic classifiers, especially when dealing with a very fine-grained set of semantic categories.
Nízká mezianotátorská shoda (IAA) je známým problémem v sémantickém značkování. IAA koreluje s granularitou lexií a oboje koreluje s množstvím informace i s její spolehlivosí. Představujeme Reliable Gain (RG), míru, která optimalizuje sémantickou granularitu se zřetelem ke spolehlivosti informace.
[ "Silvie Cinková", "Martin Holub", "Vincent Kríž" ]
https://www.semanticscholar.org/paper/12a1ac9c604635f61864be93a0d7fd19f625b85d/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
Optimizing semantic granularity for NLP - report on a lexicographic experiment
Optimalizace sémantické granularity pro NLP - zpráva o jednom lexikografickém experimentu
Experiments with semantic annotation based on the Corpus pattern Analysis and the lexical resource PDEV (Hanks and Pustejovsky, 2005), revealed a need of an evaluation measure that would identify the optimum relation between the semantic granularity of the semantic categories in the description of a verb and the reliability of the annotation expressed by the interannotator agreement (IAA). We have introduced the Reliable Information Gain (RG), which computes this relation for each tag selected by the annotators and relates it to the entry as a whole, suggesting merges of unreliable tags whenever it would increase the information gain of the entire tagset (the number of semantic categories in an entry). The merges suggested in our 19-verb sample correspond with common sense. One of the possible applications of this measure is quality management of the entries in a lexical resource.
Experimenty se sémantickou anotací založenou na Corpus Pattern Analysis a lexikálním zdroji PDEV (Hanks a Pustejovsky, 2005) ukázaly potřebu evaluační míry, která by identifikovala optimální vztah mezi sémantickou granularitou sémantických kategorií v lexikálním popisu slovesa a spolehlivostí anotace, která se měří pomocí mezianotátorské shody. Představujeme takovou míru.
[ "Silvie Cinková", "Martin Holub", "Vincent Kríž" ]
https://www.semanticscholar.org/paper/0f267b00e2024159738cccd50b8254a1ed40c6d9/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
Combining Manual and Automatic Annotation of a Learner Corpus
Kombinace ruční a automatické anotace žákovského korpusu
We present an approach to building a learner corpus of Czech, manually corrected and annotated with error tags using a complex grammar-based taxonomy of errors in spelling, morphology, morphosyntax, lexicon and style. This grammar-based annotation is supplemented by a formal classification of errors based on surface alternations. To supply additional information about non-standard or ill-formed expressions, we aim at a synergy of manual and automatic annotation, deriving information from the original input and from the manual annotation.
Článek popisuje přístu k automatické a manuální anotaci žákovského korpusu.
[ "Tomáš Jelínek", "Barbora Štindlová", "Alexandr Rosen", "Jirka Hana" ]
https://www.semanticscholar.org/paper/271fcd35496ccec10aa0080b9dee3cb966f0a4da/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
How to Measure Word Order Freedom for Natural Languages?
Jak měřit volnost slovosledu přirozených jazyků?
The paper focuses on a phenomenon of free word order through the analysis by reduction. It exploits its formal background and data types and studies the word order freedom of a (natural) language. We propose and discuss several variants of a measure based on a number of word order shifts (i.e., word order changes preserving syntactic correctness, individual word forms, their morphological characteristics and their surface dependency relations). Such measure helps to understand the difference between word order complexity (how difficult it is to parse sentences with more complex word order) and word order freedom (to which extent it is possible to change the word order without causing a change of individual word forms, their morphological characteristics and/or their surface dependency relations).
Příspěvek se zabývá volným slovosledem z pohledu redukční analýzy. Využívá jejích formálních základů a datové struktury. Navrhujeme několik variant míry volnosti slovosledu založených na operaci posouvání (shift). tedy operaci spočívající ve slovosledných změnách zachovávajících syntaktickou správnost věty. Tyto míry umožňují pochopit a studovat rozdíl mezi komplexitou slovosledu (jak složité je analyzovat věty s komplexním slovosledem) a volností slovosledu (jak je možné měnit slovosled při zachování závislostních relací ve větě).
[ "Vladislav Kuboň", "Markéta Lopatková", "Martin Plátek" ]
https://www.semanticscholar.org/paper/69eaf6f4b86716ffff28eceecb2e627d02d03480/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
On Formalization of Word Order Properties
O formalizaci slovosledných vlastností
This paper contains an attempt to formalize the degree of word order freedom for natural languages. It exploits the mechanism of the analysis by reduction and defines a measure based on a number of shifts performed in the course of the analysis. This measure helps to understand the difference between the word order complexity (how difficult it is to parse sentences with more complex word order) and word order freedom in Czech (to which extent it is possible to change the word order without causing a change of individual word forms, their morphological characteristics and/or their surface dependency relations). We exemplify this distinction on a pilot study on Czech sentences with clitics.
Tento článek se zabývá formalizací popisu volného slovosledu přirozených jazyků. Využívá mechanismu redukční analýzy a definuje míru volnosti slovosledu na základě počtu přesunů provedených v průběhu analýzy. Tato míra umožňuje rozlišit složitost slovosledu (jak obtížné je analyzovat věty se složitějším slovosledem) a volnost slovosledu (do jaké míry je možné měnit slovosled, aniž by došlo ke změně jednotlivých slovních tvarů, jejich morfologické charakteristiky a / nebo jejich povrchově syntaktických vztahů). Tento rozdíl je ilustrován na pilotní studii českých vět s klitikami.
[ "Vladislav Kuboň", "Markéta Lopatková", "Martin Plátek" ]
https://www.semanticscholar.org/paper/1e4bead0ee9b156f02b1104eb248806a2ea023c7/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
Studying Formal Properties of a Free Word Order Language
Příspěvek ke studiu formálních vlastností jazyků s volným slovosledem
The paper investigates a phenomenon of free word order through the analysis by reduction. It exploits its formal background and data types and studies the word order freedom by means of the minimal number of word order shifts (word order changes preserving syntactic correctness, individual word forms, their morphological characteristics and/or their surface dependency relations). The investigation focuses upon an interplay of two phenomena related to word order: (non-)projectivity of a sentence and number of word order shifts within the analysis by reduction. This interplay is exemplified on a sample of Czech sentences with clitics.
Příspěvek zkoumá fenomén volného slovosledu přirozeného jazyka metodou redukční analýzy. Využívá přitom formální rámec a datový typ této analytické metody a uplatňuje ho na zkoumání minimálního počtu slovosedných posunů (tj. slovosledných změn zachovávajících syntaktickou správnost, jednotlivé slovní tvary, jejicj morfologickou a syntaktickou charakteristiku). Zkoumání se soustřeďují na dva vzájemně se ovlivňující jevy související se slovosledem: (ne)projektivitu a počet přesunů. Tyto jevy jsou ilustrovány na vzorlu českých vět obsahujících klitiky.
[ "Vladislav Kuboň", "Markéta Lopatková", "Martin Plátek" ]
https://www.semanticscholar.org/paper/d8941c5101212cf316a565287dd227eed427eec1/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
Machine Translation Among Related Slavic Languages
Strojový překlad příbuzných slovanských jazyků
This chapter describes the history of experiments in the field of machine translation among related Slavic languages, methods used and results achieved.
Kapitola popisuje historii pokusů s automatickým překladem mezi příbuznými jazyky, použité metody a dosažené výsledky.
[ "Vladislav Kuboň", "Petr Homola" ]
https://www.semanticscholar.org/paper/43d31d4cff9beb3f0f455e4b783f9bda26f19dc9/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
Prague Dependency Treebank 2.5 -- a revisited version of PDT 2.0
Pražský závislostní korpus 2.5 -- rozšířená verze PDT 2.0
We present the Prague Dependency Treebank 2.5, the newest version of PDT and the first to be released under a free license. We show the benefits of PDT 2.5 in comparison to other state-of-the-art treebanks. We present the new features of the 2.5 release, how they were obtained and how reliably they are annotated. We also show how they can be used in queries and how they are visualised with tools released alongside the treebank.
Představujeme nejnovější verzi Pražského závislostního treebanku PDT 2.5, který bude poprvé vydán pod veřejnou licencí. Výhody PDT 2.5 ukážeme na srovnání s nejmodernějšími treebanky. Představíme nové vlastnosti verze 2.5, popíšeme, jak byly anotovány i jak spolehlivá tato anotace je. Ukážeme, jakými dotazy lze nové jevy hledat a jak se zobrazují v nástrojích dodávaných spolu s treebankem.
[ "Eduard Bejček", "Jarmila Panevová", "Jan Popelka", "Pavel Straňák", "Magda Ševčíková", "Jan Štěpánek", "Zdeněk Žabokrtský" ]
https://www.semanticscholar.org/paper/577119914a839521c7d134960968c5f13f1f46ca/
null
null
null
null
null
null
null
null
null
null
null
null
cs
2012
Predicative adverbs in the grammatical description of Czech
Predikativum v gramatickém popisu češtiny
From a number of Czech adjectives, both an adverb with the suffix -o and an adverb with the suffix -e are derived, cf. deštivý `rainy’ – deštivo/deštivě, drahý `expensive’ – draho/draze. These pairs of adverbs have often the same meaning but the adverb with the suffix -o occurs as a part of the predicate (with the verb být `to be’; ráno bylo deštivo `in the morning it was rainy’) while the adverb with the suffix -e is often specialized for the adverbial function (e.g. ráno vypadalo deštivě/*deštivo `the morning looked rainy’). On the functional specialization Komárek’s (1954) proposal to separate the adverbs with the suffix -o as an autonomous part of speech was based. Since the actual corpus material indicates that both the adverb with the suffix -o and the adverb with -e are acceptable in the same context (cf. je tam draho/draze `it is an expensive place to live’, draho/draze prodat `to sell dear’ VS. přišlo ho to draho/*draze ` it cost him dear’, draze/*draho za to zaplatil `he paid for this’) but the (non)acceptability cannot be probably explained by a grammatical principle, it seems to be appropriate to consider these words further as adverbs and to describe them as separate lexical units rather than a homogenous group.
Od řady českých adjektiv se tvoří jak adverbium s příponou -e, tak adverbium s příponou -o, př. deštivý – deštivě/deštivo, drahý – draze/draho. Deadjektivní adverbia s příponou -o často vystupují jako součást verbonominálního predikátu se slovesem být v neosobních konstrukcích, některá z nich se na tuto větněčlenskou funkci dokonce omezují (př. na horách bylo deštivo), jiná plní kromě této funkce i funkce jiné (srov. je tam draho vs. přišlo ho to draho). Predikativní funkce deadjektivních adverbií na -o se stala hlavním argumentem pro úvahy o vyčlenění těchto adverbií (spolu s dalšími výrazy) jako samostatného slovního druhu, tzv. predikativ. V příspěvku probíráme syntaktické a sémantické vlastnosti těchto adverbií na -o na základě aktuálního korpusového materiálu. Jako výhodnější se ukazuje popisovat tato adverbia jako samostatné lexikální jednotky, nikoli jako skupinu.
[ "Magda Ševčíková" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
Sentence Modality Assignment in the Prague Dependency Treebank
Určování větné modality v Pražském závislostním korpusu
The paper focuses on the annotation of sentence modality in the Prague Dependency Treebank (PDT). Sentence modality is expressed by a combination of several means in Czech, from which the category of verbal mood and the final punctuation of the sentence are the most important ones. In PDT 2.0, sentence modality was assigned semi-automatically to the root node of each sentence (tree) and further to the roots of parenthesis and direct speech subtrees. As this approach was too simple to adequately represent the linguistic phenomenon in question, the method for assigning the sentence modality has been revised and elaborated for the forthcoming version of the treebank (PDT 3.0).
Článek se zabývá anotací větné modality v Pražském závislostním korpusu (PDT). Větná modalita je v češtině vyjádřena kombinací několika faktorů, především slovesným způsobem a koncovou interpunkcí. V PDT 2.0 byla větná modalita přiřazena poloautomaticky kořeni každé věty (stromu) a dále kořenům stromů reprezentujících vsuvku nebo přímou řeč. Tento přístup byl příliš zjednodušující pro adekvátní zachycení daného jevu, proto byla metoda přiřazení větné modality pro příští vydání treebanku (PDT 3.0) zrevidována a rozpracována.
[ "Magda Ševčíková", "Jiří Mírovský" ]
https://www.semanticscholar.org/paper/d1ba9814b2149c2103236a8039ba6b1fdd49d566/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
Reinforcement learning for spoken dialogue systems using off-policy natural gradient method
Zpětnovazební učení pro hlasové dialogové systémy používající off-policy gradientní metody
Reinforcement learning methods have been successfully used to optimise dialogue strategies in statistical dialogue systems. Typically, reinforcement techniques learn on-policy i.e., the dialogue strategy is updated online while the system is interacting with a user. An alternative to this approach is off-policy reinforcement learning, which estimates an optimal dialogue strategy offline from a fixed corpus of previously collected dialogues. This paper proposes a novel off-policy reinforcement learning method based on natural policy gradients and importance sampling. The algorithm is evaluated on a spoken dialogue system in the tourist information domain. The experiments indicate that the proposed method learns a dialogue strategy, which significantly outperforms the baseline handcrafted dialogue policy
Zpětno vazební učení již bylo úspěšně použito k optimalizaci statisktických dialogových systémů. Typicky zpětnovazební učení se učí online on-policy tj. v přímé interakci s uživatelem. Alternativou k tomuto přístupu je off-policy učení kdy otimální strategie řízení je určena z korpusu již dříve pořízených dialogů. Tento článek prezentuje a nový zpětnovazební algoritmus založený na přirozených gradientech a vhodné adaptaci samplování dat. Experimenty ukazují, že prezentovaný algoritmus je schopen se naučit strategii řízení, která je lepší než manuálně vytvořená strategie řízení.
[ "Filip Jurčíček" ]
https://www.semanticscholar.org/paper/d7523ad4223b600398d11538256611531f84d098/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
Dependency Relations Labeller for Czech
Značkovač závislostních vztahů pro češtinu
We present a MIRA-based labeller designed to assign dependency relation labels to edges in a dependency parse tree, tuned for Czech language. The labeller was created to be used as a second stage to unlabelled dependency parsers but can also improve output from labelled dependency parsers. We evaluate two existing techniques which can be used for labelling and experiment with combining them together. We describe the feature set used. Our final setup significantly outperforms the best results from the CoNLL 2009 shared task.
Představujeme značkovač závislostních vztahů založený na metodě MIRA, vyladěný na češtině.
[ "Rudolf Rosa", "David Mareček" ]
https://www.semanticscholar.org/paper/caca4ea7e88fc0cbc77e93cdda8cf40189f9c192/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
DEPFIX: A System for Automatic Correction of Czech MT Outputs
DEPFIX: Systém pro automatickou opravu českých výstupů MT
We present an improved version of DEPFIX, a system for automatic rule-based post-processing of English-to-Czech MT outputs designed to increase their fluency. We enhanced the rule set used by the original DEPFIX system and measured the performance of the individual rules. We also modified the dependency parser of McDonald et al. (2005) in two ways to adjust it for the parsing of MT outputs. We show that our system is able to improve the quality of the state-of-the-art MT systems.
Prezentujeme vylepšenou verzi systému DEPFIX pro automatický post-editing výstupu z anglicko-českého strojového překladu, který se snaží zlepšovat jeho plynulost. Rozšířili jsme sadu pravidel použitou původním systémem DEPFIX a změřili výkon jednotlivých pravidel. Navíc jsme dvěma způsoby upravili parser McDonalda et al. (2005) pro zvýšení kvality parsingu výstupu strojového překladu. Ukazujeme, že náš systém je schopný zlepšit výstup nejmodernějších překladových systémů.
[ "Rudolf Rosa", "David Mareček", "Ondřej Dušek" ]
https://www.semanticscholar.org/paper/3be210bb67fa9d080eeba387cf3cf73a665c2c00/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
Using Parallel Features in Parsing of Machine-Translated Sentences for Correction of Grammatical Errors
Využití paralelních rysů v parsingu strojově přeložených vět pro korekci gramatických chyb
In this paper, we present two dependency parser training methods appropriate for parsing outputs of statistical machine translation (SMT), which pose problems to standard parsers due to their frequent ungrammaticality. We adapt the MST parser by exploiting additional features from the source language, and by introducing artificial grammatical errors in the parser training data, so that the training sentences resemble SMT output. We evaluate the modified parser on DEPFIX, a system that improves English-Czech SMT outputs using automatic rule-based corrections of grammatical mistakes which requires parsed SMT output sentences as its input. Both parser modifications led to improvements in BLEU score; their combination was evaluated manually, showing a statistically significant improvement of the translation quality.
Představujeme dvě metody pro trénování závislostního parseru vhodného pro parsing výstupů strojového překladu. Upravili jsme MST parser použitím dalších rysů ze zdrojového jazyka a zavedením umělých gramatických chyb do trénovacích dat parseru, takže tato více odpovídají výstupu storjového překladu. Upravený parser evaluujeme na systému DEPFIX, který zlepšuje výstupy anglicko-českého strojového překladu automatickými opravami založenými na pravidlech. Obě úpravy parseru vedou ke zvýšení skóre BLEU; jejich kombinace byla evaluována manuálně a vykazuje statisticky signifikantní zlepšení kvality překladu.
[ "Rudolf Rosa", "Ondřej Dušek", "David Mareček", "Martin Popel" ]
https://www.semanticscholar.org/paper/bb4c579db6825727006e9879b11eb2adef206fa6/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
Introducing context lexical models to TectoMT
Zavádění kontextových lexikálních modelů do TectoMT
Much of the attention in the MT community has recently been devoted to improving the lexical choice in the target language by capturing a context wider than just a single sentence. In this talk I will present our contribution to these efforts, particularly the progress in enriching translation models for English to Czech translation within the TectoMT system. First of all, we performed a pure technical step. We replaced the MaxEnt learning module used so far with the Vowpal Wabbit learning toolkit. It not only allows us to train our models faster, to exploit more training examples and features, but it also offers rich parametrization options, which can together lead to improvement of our MT system. However, the main objective of this ongoing work is to explore the potential of lexical context features to improve the translation. We do so by including the standard bag-of-words features and by introducing novel features representing concepts coming from Explicit Semantic Analysis, which was originally developed in the field of Information Retrieval.
Množstvo pozornosti v MT komunitě bylo v nedávne době věnováno zlepšování lexikálního výběru v cílovém jazyce pomocí zachycení kontextu širšího nežli jedna věta. V této přednášce prezentuji náš příspěvek k těmto snahám, konkrétně pokrok v obohacování překladových modelů pro překlad z angličtiny do češtiny systémem TectoMT. Ze všeho nejdřív jsme provedli čistě techický krok. Nahradili jsme doteď používaný modul MaxEnt na strojové učení nástrojem Vowpal Wabbit. Ten nám nejenom umožňuje trénovat modely rychleji a tím využít víc trénovacích příkladů a rysů, ale také nabízí bohaté možnosti parametrizace, co dohromady může vést k zlepšení našeho MT systému. Nicméně, hlavním cílem tohoto pořád běžícího projektu je prozkoumat potenciál lexikálních kontextových rysů na vylepšení překladu. Činíme tak přidáním standardních bag-of-words rysů a zavedením nových rysů reprezentujících koncepty z Explicitní Sémantické Analýzy, co je metoda původně vyvynuta v oboru Dobývaní znalostí.
[ "Michal Novák" ]
https://www.semanticscholar.org/paper/1aeabdf04096d4eb0e0f4e7022e3fd1ce064ddce/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
Utilization of context in Machine Translation
Využití kontextu ve strojovém překladu
In this talk I presented the shift of my research from utilization of coreference to utilization of text context in general in the task of Machine Translation. I introduced lexical discriminative translation models and the initial experiments with them integrated within the English to Czech translation scenario in TectoMT system.
V této přednášce jsem prezentoval posun témy mojeho výzkumu z využití koreference na využití textového kontextu všeobecně v úloze strojového překladu. Představil jsem lexikální diskriminativní překladové modely a prvotní experimenty s nimi v rámci překladu z angličtiny do češtiny systémem TectoMT.
[ "Michal Novák" ]
https://www.semanticscholar.org/paper/c2745e86ecc9bec372690cced53ccfdf44f407f8/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
Yet Another Language Identifier
Ještě jeden rozpoznávač jazyků
Language identification of written text has been studied for several decades. Despite this fact, most of the research is focused on a few most spoken languages, whereas the minor ones are ignored. The identification of a larger number of languages brings new difficulties that do not occur for a few languages. These difficulties are causing decreased accuracy. The objective of this paper is to investigate the sources of such degradation. In order to isolate the impact of individual factors, 5 different algorithms and 3 different number of languages are used. The Support Vector Machine algorithm achieved an accuracy of 98% for 90 languages and the YALI algorithm based on a scoring function had an accuracy of 95.4%. The YALI algorithm has slightly lower accuracy but classifies around 17 times faster and its training is more than 4000 times faster. Three different data sets with various number of languages and sample sizes were prepared to overcome the lack of standardized data sets. These data sets are now publicly available.
Identifikace jazyku psaného textu se zkoumá již několik desetiletí. Navzdory tomuto faktu se většina vůzkumů soustředila pouze na pár nejčastěji používaných jazyků zatímco ty ostatní jsou ignorovány. Při identifikaci velkého množství jazyků je nutné řešit jiné problémy než u identifikace malého množství jazyků, protože v opačném případě nastává pokles přesnosti. Cílem tohoto článku je prozkoumat důvody tohoto poklesu. Aby bylo možné izolovat jednotlivé faktory použili jsme 5 různých algoritmů a 3 různé počty jazyků. SVM algoritmus dosáhl úspěšnosti 98% pro 90 jazyků a YALI algoritmus založená na ohodnocující funkci dosáhl úspěšnosti 95,4%. YALI algoritmus je sice nepatrně horší, ale jazyky identifikuje 17x rychleji a jeho trénování je dokonce 4000x rychlejší. Připravili jsme také 3 různé datasety s různým počtem jazyků a velikostí vzorků, abychom překonali nedostatek veřejně dostupných trénovacích dat.
[ "Martin Majliš" ]
https://www.semanticscholar.org/paper/0e4d3db0f2a14eb0d76de2f37a544b03bddcc75a/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
Language Richness of the Web
Bohatství jazyků na webu
We have built a corpus containing texts in 106 languages from texts available on the Internet and on Wikipedia. The W2C Web Corpus contains 54.7 GB of text and the W2C Wiki Corpus contains 8.5 GB of text. The W2C Web Corpus contains more than 100 MB of text available for 75 languages. At least 10 MB of text is available for 100 languages. These corpora are a unique data source for linguists, since they outclass all published works both in the size of the material collected and the number of languages covered. This language data resource can be of use particularly to researchers specialized in multilingual technologies development. We also developed software that greatly simplifies the creation of a new text corpus for a given language, using text materials freely available on the Internet. Special attention was given to components for filtering and de-duplication that allow to keep the material quality very high.
Vytvořili jsme korpus obsahující texty ve 106 jazycích z dokumentů, které jsou dostupné na Internetu a Wikipedii. W2C Web Corpus obsahuje 54,7 GB textu a W2C Wiki Corpus obsahuje 8,5 GB textu. W2C Web Corpus obsahuje více než 100 MB textu pro 75 jazyů a alespoň 10 MB textu pro 100 jazyků. Tyto korpusy jsou jedinečným zdrojem dat pro lingvistiku, protože překonávají všechny dosud publikované práce, jak v množství nashromážděných textů, tak i v množství obsažených jazyků. Tento zdroj dat může být především užitečný pro vědce specializujícící se na vývoj vícejazyčných technologií. Také jsme vyvinuli software, který výrazně usnadňje tvorbu korpusů pro libovolný jazyk z textů volně dostupných na Internetu. Při vývoji jsme se hlavně zaměřili na komponenty pro filtrovaní a odstraňování duplicit, což nám umožnilo dosáhnout vysoké kvality výsledných dat.
[ "Martin Majliš", "Zdeněk Žabokrtský" ]
https://www.semanticscholar.org/paper/a8ede4e330b8931011004b3a961857abff378dde/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
Morphological Processing for English-Tamil Statistical Machine Translation
Morfologické zpracování pro statistický strojový překlad z angličtiny do tamilštiny
Various experiments from literature suggest that in statistical machine translation (SMT), applying either pre-processing or post-processing to morphologically rich languages leads to better translation quality. In this work, we focus on the English-Tamil language pair. We implement suffix-separation rules for both of the languages and evaluate the impact of this preprocessing on translation quality of the phrase-based as well as hierarchical model in terms of BLEU score and a small manual evaluation. The results confirm that our simple suffix-based morphological processing helps to obtain better translation performance. A by-product of our efforts is a new parallel corpus of 190k sentence pairs gathered from the web.
Různé experimenty z literatury naznačují, že ve statistickém strojového překladu (SMT), předzpracování nebo závěrečné úpravy pro morfologicky bohaté jazyky vedou k lepší kvalitě překladu. V této práci jsme se zaměřili na jazykový pár angličtina-tamilština.
[ "Loganathan Ramasamy", "Ondřej Bojar", "Zdeněk Žabokrtský" ]
https://www.semanticscholar.org/paper/22b5cac3c84cbaf7fee4291e3c862f1d515fca81/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
Prague Dependency Style Treebank for Tamil
Pražský závislostní korpus styl pro Tamil
Annotated corpora such as treebanks are important for the development of parsers, language applications as well as understanding of the language itself. Only very few languages possess these scarce resources. In this paper, we describe our efforts in syntactically annotating a small corpora (600 sentences) of Tamil language. Our annotation is similar to Prague Dependency Treebank (PDT) and consists of annotation at 2 levels or layers: (i) morphological layer (m-layer) and (ii) analytical layer (a-layer). For both the layers, we introduce annotation schemes i.e. positional tagging for m-layer and dependency relations for a-layers. Finally, we discuss some of the issues in treebank development for Tamil.
Anotované korpusy jako treebanks jsou důležité pro vývoj analyzátorů, jazykové aplikace, stejně jako porozumění Jazyk sám. Jen velmi málo jazyků mají tyto omezené zdroje. V tomto článku si popíšeme naše úsilí v syntakticky anotace malé korpusy (600 vět) z Tamil jazyce. Naše anotace je podobný Pražského závislostního korpusu (PDT) a skládá se z Anotace na 2 podlažích či vrstev: (i) Morfologická rovina (m-layer) a (ii) analytické vrstvy (vrstvy). U obou vrstev, uvádíme anotace programů, tj. poziční značení pro m-layer a vztahy závislosti na několika vrstev. Na závěr budeme diskutovat některé otázky v korpus vývoj pro Tamil.
[ "Loganathan Ramasamy", "Zdeněk Žabokrtský" ]
https://www.semanticscholar.org/paper/77ef6c449d3b7f5f5c55a06564b79eb4438c85b9/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
The Study of Effect of Length in Morphological Segmentation of Agglutinative Languages
STUDIUM VLIVU délky v morfologických segmentaci aglutinační JAZYKŮ
Morph length is one of the indicative feature that helps learning the morphology of languages, in particular agglutinative languages. In this paper, we introduce a simple unsupervised model for morphological segmentation and study how the knowledge of morph length affect the performance of the segmentation task under the Bayesian framework. The model is based on (Goldwater et al., 2006) unigram word segmentation model and assumes a simple prior distribution over morph length. We experiment this model on two highly related and agglutinative languages namely Tamil and Telugu, and compare our results with the state of the art Morfessor system. We show that, knowledge of morph length has a positive impact and provides competitive results in terms of overall performance.
Morph délka je jedním z orientační funkce , která pomáhá učit morfologii jazyků, zejména aglutinační jazyky. V tomto článku vám představíme jednoduchý bez dozoru model pro morfologické segmentace a pozorujete, jak se znalosti morph Délka ovlivnit výkon segmentaci úkol v rámci bayesovského rámce. Model je založen na (Goldwater et al., 2006) unigram slovo segmentace modelu a předpokládá jednoduchou distribuci přes předchozí morph délka. Jsme experimentovat tento model na dvou velmi příbuzných a aglutinační jazyky Tamil a Telugu to, a porovnávat naše výsledky s nejmodernější Morfessor systém. Ukazujeme, že znalost morph délka má pozitivní vliv a poskytuje konkurenceschopných výsledků, pokud jde o celkový výkon.
[ "Loganathan Ramasamy", "Zdeněk Žabokrtský", "Sowmya Vajjala" ]
https://www.semanticscholar.org/paper/e6bffa8c2b39f57ab9a614be2f442c95bcb7305f/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
Retraining Machine Translation with Post-edits to Increase Post-editing Productivity in Content Management Systems
Retrénování strojového překladu s post-editacemi ke zlepšení post-editační produktivity v systémech pro správu obsahu
The post-editing of machine translated content has proved to be more productive than translating from scratch in the localisation industry. A study carried out at Autodesk shows that Machine Translation (MT) and post-editing of technical documentation by professional translators provides a sustained productivity increase across different languages.
Tato práce se zabývá využitím manuálně post-editovaných automatických překladů pro iterativní trénování systémů statistického strojového překladu.
[ "Antonio Toral", "Leroy Finn", "Dominic Jones", "Pavel Pecina", "David Lewis", "Declan Groves" ]
null
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
Efficiency-based Evaluation of Aligners for Industrial Applications
Vyhodnocení efektivity zarovnávačů pro průmyslové aplikace
This paper presents a novel efficiencybased evaluation of sentence and word aligners. This assessment is critical in order to make a reliable use in industrial scenarios. The evaluation shows that the resources required by aligners differ rather broadly. Subsequently, we establish limitation mechanisms on a set of aligners deployed as web services. These results, paired with the quality expected from the aligners, allow providers to choose the most appropriate aligner according to the task at hand.
Tato práce se zabývá evaluací výkonosti větných a slovních zarovnávačů, což je velice významný aspekt pro průmyslové nasazení, a ukazuje, že zdroje vyžadované různými nástroji se velice liší.
[ "Antonio Toral", "Marc Poch", "Pavel Pecina", "Gregor Thurmair" ]
https://www.semanticscholar.org/paper/d3129dbd09cc4ea9f9497637b15571f6d892aab6/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
TerrorCat: a Translation Error Categorization-based MT Quality Metric
TerrorCat: metrika hodnocení kvality překladu založená na kategorizaci chyb
We present TerrorCat, a submission to the WMT’12 metrics shared task. TerrorCat uses frequencies of automatically obtained translation error categories as base for pairwise comparison of translation hypotheses, which is in turn used to generate a score for every translation. The metric shows high overall correlation with human judgements on the system level and more modest results on the level of individual sentences.
Představujeme TerrorCat, metriku kvality strojového překladu s níž jsme se účastnili soutěže WMT12. TerrorCat se opírá o automaticky odhadnuté počty několika kategorií překladových chyb a předpovídá tak, která ze dvou kandidátských vět je lepší.
[ "Mark Fishel", "Rico Sennrich", "Maja Popović", "Ondřej Bojar" ]
https://www.semanticscholar.org/paper/6acdc433d1f4d93fc971f287df60aa8fab10d3a1/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
Terra: a Collection of Translation Error-Annotated Corpora
Terra: sbírka korpusů s vyznačenými chybami v překladu
Recently the first methods of automatic diagnostics of machine translation have emerged; since this area of research is relatively young, the efforts are not coordinated. We present a collection of translation error-annotated corpora, consisting of automatically produced trans- lations and their detailed manual translation error analysis. Using the collected corpora we evaluate the available state-of-the-art methods of MT diagnostics and assess, how well the methods perform, how they compare to each other and whether they can be useful in practice.
V nedávné době se objevily první automatické metody identifikace chyb v překladu. Představujeme materiál, na němž lze takové metody vyhodnocovat, konkrétně kolekci textů, kde lidé ručně překladové chyby označili.
[ "Mark Fishel", "Ondřej Bojar", "Maja Popović" ]
https://www.semanticscholar.org/paper/a709b0bf733b607a9fd95a07c1e5e899e5a25063/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
Automatic MT Error Analysis: Hjerson Helping Addicter
Automatická analýza chyb strojového překladu: Hjerson pomáhá Addicteru
We present a complex, open source tool for detailed machine translation error analysis providing the user with automatic error detection and classification, several monolingual alignment algorithms as well as with training and test corpus browsing. The tool is the result of a merge of automatic error detection and classification of Hjerson (Popović, 2011) and Addicter (Zeman et al., 2011) into the pipeline and web visualization of Addicter. It classifies errors into categories similar to those of Vilar et al. (2006), such as: morphological, reordering, missing words, extra words and lexical errors. The graphical user interface shows alignments in both training corpus and test data; the different classes of errors are colored. Also, the summary of errors can be displayed to provide an overall view of the MT system’s weaknesses. The tool was developed in Linux, but it was tested on Windows too.
Představujeme komplexní open-source nástroj pro detailní analýzu chyb strojového překladu, který umožňuje automatickou detekci a klasifikaci chyb, poskytuje několik algoritmů pro jednojazyčné zarovnání (alignment) slov a procházení trénovacích a testovacích korpusů.
[ "Jan Berka", "Ondřej Bojar", "Mark Fishel", "Maja Popović", "Daniel Zeman" ]
https://www.semanticscholar.org/paper/c630c9c709197684b3024d23baefaeb30c47e4ad/
null
null
null
null
null
null
null
null
null
null
null
null
en
2012
On a Corpus of Older Czech Texts and Its Usage
Ke korpusu starší češtiny a jeho využití
The article describes the Corpus of older Czech texts and demonstrates how it is possible to use it in linguistics.
Článek popisuje Korpus starších českých textů a ukazuje možnosti jeho využití v lingvistice.
[ "František Martínek", "Kateřina Rysová" ]
null
null
null
null
null
null
null
null
null
null
null
null
null