Претрага

Упит dcterms_subject_txt:literary corpus

83 items

  • FrameNet Lexical Database: Presenting a Few Frames Within the Risk Domain

    Aleksandra Marković, Ranka Stanković, Natalija Tomić, Olivera Kitanović (2021)
    U radu se daje kratak prikaz teorije semantike okvira, na kojoj je zasnovana leksička baza Frejmnet. Predstavljena je koncepcija ove mreže, kao i mogućnosti njene primene. Predstavljena je i leksička analiza koja se primenjuje u projektu izrade Frejmneta i ukazano na razlike između analize zasnovane na okviru u odnosu na analizu zasnovanu na reči. Zatim je prikazano nekoliko povezanih okvira koje prizivaju reči iz domena rizika. U radu je predstavljena i platforma NLTК pomoću koje se mogu koristiti ...
    Srpski jezik, semantika okvira, FrameNet, scenario rizika, rudarski korpus, obrada prirodnog jezika
    ... (Natural Language Toolkit) suite, which provides a good natural language pro- cessing resource. The last chapter shows a corpus search of the noun risk in a mining- themed corpus. We also present its most common collocates, word sketch, individual pattern concordances, thesaurus entry of its synonyms ...
    ... analysis of the meaning of an LU, its lexical surroundings, phrases and grammatical constructions in which it appears in the corpus, the context in which it is used provided by corpus examples, as well as all the phrases in which the LU fulfills its full semantic potential. This approach consists of listing ...
    ... 2016, 21) and its use is looked into by extracting sentences, which contain it, from the corpus. A lexicographer working in FrameNet compares his or her insight into the meaning of a target lexeme, based on corpus examples, to the meaning given in descriptive dictionaries.8 Once he gets a clearer idea ...

  • SrpELTeC: A Serbian Literary Corpus for Distant Reading

    Ranka Stanković, Cvetana Krstev, Duško Vitas (2024)
    U članku je predstavljen SrpELTeC, korpus razvijen u okviru akcije COST Distant Reading for European Literary History (CA16204). Svi romani u SrpELTeC-u su odabrani, pripremljeni i obeleženi korišćenjem zajedničkih principa uspostavljenih za sve jezičke zbirke u Evropskoj zbirci književnog teksta (ELTeC). Navedeni su izazovi i rešenja u pripremi SrpELTeC od nule. Svi romani su ručno kodirani u TEI sa bogatim metapodacima i strukturnim napomenama. Automatska anotacija je uključivala POS-označavanje, lematizaciju i imenovane entitete, oslanjajući se na resurse za obradu ...
    digital humanities, Serbian literature, text corpora, distant reading , linked data, named entity recognition, text analytics

  • Corpus-based bilingual terminology extraction in the power engineering domain

    Tanja Ivanović, Ranka Stanković, Branislava Šandrih Todorović, Cvetana Krstev (2022)
    Ovaj rad predstavlja resurse i alate koji se koriste za ekstrkciju i evaluaciju dvojezične, englesko-srpske terminologije u domenu energetike. Resursi se sastoje od postojeće opšte i domenske leksike i domenskog paralelnog korpusa; alati uključuju ekstraktore termina za oba jezika i alat za poravnavanje segmenata koji pripadaju korpusnim rečenicama. Sistem je testiran variranjem funkcije podudaranja koja utvrđuje prisustvo ekstrahovanog termina u poravnatom segmentu (odsečak), u rasponu od veoma labavog do strogog. Procena rezultata je pokazala da je preciznost izdvajanja termina ...
    Library and Information Sciences, Communication, Language and Linguistics

  • Bilingual lexical extraction based on word alignment for improving corpus search

    Jelena Andonovski, Branislava Šandrih, Olivera Kitanović (2019)
    Library and Information Sciences,Computer Science Applications

  • Machine Learning and Deep Neural Network-Based Lemmatization and Morphosyntactic Tagging for Serbian

    Ranka Stanković, Branislava Šandrih, Cvetana Krstev, Miloš Utvić, Mihailo Škorić (2020)
    The training of new tagger models for Serbian is primarily motivated by the enhancement of the existing tagset with the grammatical category of a gender. The harmonization of resources that were manually annotated within different projects over a long period of time was an important task, enabled by the development of tools that support partial automation. The supporting tools take into account different taggers and tagsets. This paper focuses on TreeTagger and spaCy taggers, and the annotation schema alignment ...
    Part-of-Speech tagging, lemmatization, corpus, evaluation, Serbian, morphological dictionary
    ... MULTEXT-East and Universal Part-of-Speech tagset. The trained models will be used to publish the new version of the Corpus of Contemporary Serbian as well as the Serbian literary corpus. The performance of developed taggers were compared and the impact of training set size was investigated, which resulted ...
    ... tagging will be used to tag the Serbian part of ELTeC (European Literary Text Collec- tion) corpus.10 It should be noted that within this action the textometry tool TXM (Heiden, 2010) is used for many 10The current version of ELTeC corpus is available at https://zenodo.org/communities/eltec/. 3961 ...
    ... History), while in one text (Novels) values of the grammatical category gender were added. 3ELTeC is a corpus prepared in the scope of COST Ac- tion CA16204 Distant Reading for European Literary History, https://www.distant-reading.net/. 3956 PoS-tag UPoS SMD TT11 MTE ADJ ✓ ✓ A ✓ A ✓ A ADP ✓ ...

  • Distant Reading in Digital Humanities: Case Study on the Serbian Part of the ELTeC Collection

    Ranka Stanković, Cvetana Krstev, Branislava Šandrih Todorović, Duško Vitas, Mihailo Škorić, Milica Ikonić Nešić (2022)
    In this paper we present the Serbian part of the ELTeC multilingual corpus of novels written in the time period 1840-1920. The corpus is being built in order to test various distant reading methods and tools with the aim of re-thinking the European literary history. We present the various steps that led to the production of the Serbian sub-collection: the novel selection and retrieval, text preparation, structural annotation, POS-tagging, lemmatization and named entity recognition. The Serbian sub-collection was published ...
    Corpus, Distant Reading, Digital Humanities, Linked Data, Named Entity Recognition, Text Analytics

  • SASA Dictionary as the Gold Standard for Good Dictionary Examples for Serbian

    Ranka Stanković, Branislava Šandrih, Rada Stijović, Cvetana Krstev, Duško Vitas, Aleksandra Marković (2019)
    У овом раду представљамо модел за избор добрих примера за речник српског језика и развој иницијалних компоненти модела. Метода која се користи заснива се на детаљној анализи различитих лексичких и синтактичких карактеристика у корпусу састављених од примера из пет дигитализованих свезака речника САНУ. Почетни скуп функција био је инспирисан сличним приступом и за друге језике. Дистрибуција карактеристика примера из овог корпуса упоређује се са карактеристиком дистрибуције узорака реченица ексцерпираних из корпуса који садрже различите текстове. Анализа је показала да ...
    Српски, добри примери из речника, аутоматизација израде речника, издвајање својстава, Машинско учење
    ... supported by the corpus data. It is common for lexicographers to look for examples in the corpus of contemporary Serbian (SrpKor, developed by D. Vitas and a group of collaborators from University of Belgrade, http://www.korpus.matf.bg.ac.rs/korpus/), which is being used as a control corpus, but they rarely ...
    ... to the corpus made of examples, we prepared a control dataset derived from various texts, which was used as a sample corpus for dictionary example extraction. The control dataset of example candidates was obtained from the digital library Biblisha8 (Stanković et al., 2017), SrpKor – the corpus of c ...
    ... syntactic features in a corpus compiled of examples from the five digitized volumes of the Serbian Academy of Sciences and Arts (SASA) dictionary. The initial set of features was inspired by a similar approach for other languages. The feature distribution of examples from this corpus is compared with the ...

  • Extraction of Bilingual Terminology Using Graphs, Dictionaries and GIZA++

    Branislava Šandrih, Ranka Stanković (2020)
    U nauci, industriji i mnogim istraživačkim oblastima, terminologija se brzo razvija. Najčešće, jezik koji je „lingua franca“ za većinu ovih oblasti je engleski. Kao posledica toga, za mnoga polja termini domena su koncipirani na engleskom, a kasnije se prevode na druge jezike. U ovom radu predstavljamo pristup za automatsko izdvajanje dvojezične terminologije za englesko-srpski jezički par koji se oslanja na usaglašeni dvojezični korpus domena, ekstraktor terminologije za ciljni jezik i alat za usklađivanje delova. Ispitujemo performanse metode na domenu ...
    ekstrakcija terminologije, validacija terminologije, GIZA++, grafovi, Unitex, klasifikacija teksta
    ... obtaining 8 different experimental settings: 1. The input domain aligned corpus (Input i) consists of: (a) the aligned corpus LIS-corpus; (b) the aligned corpus LIS-corpus extended with the bilingual aligned pairs bi-list (LIS-corpus+); 2. The list of domain terms for the source language (Input ii) is ...
    ... lexicon for a specific domain, we combined and compared several settings. Besides using only a parallel sentence-aligned corpus, we conducted an experiment where sentences from the corpus were extended with a bilingual list of inflected word forms from a general-purpose dictionary, similarly as in (Tsvetkov ...
    ... relies on several lexical resources and tools: i A sentence-aligned domain-specific corpus involving a source and a target language, denoted as S(text.align) ↔ T (text.align). In this paper we refer to this tool as LIS-corpus. As a textual resource, twelve issues with a total of 84 papers were aligned at ...

  • Towards ELTeC-LLOD: European Literary Text Collection Linguistic Linked Open Data

    Ranka Stanković, Christian Chiarcos, Miloš Utvić, Olivera Kitanović (2023)
    Овај рад описује студију случаја о генерисању повезаних података креираних на основу обечежених текстуалних корпуса коришћењем формата размене података у обради природних језика (NIF). Као основа за ово истраживање послужио је подскуп корпуса ELTeC, који се састоји од 900 романа из периода 1840-1920 за 9 европских језика. Верзија романа са коментарима, у такозваном TEI level-2 формату, трансформисана је у NIF, формат заснован на RDF/OWL који има за циљ постизање интероперабилности између алата за обраду природних језика, језичких ресурса и ...
    повезани отворени подаци, корпус, SrpELTeC, NIF

  • OntoLex Publication Made Easy: A Dataset of Verbal Aspectual Pairs for Bosnian, Croatian and Serbian

    Ranka Stanković, Maxim Ionov, Medina Bajtarević, Lorena Ninčević (2024)
    Ovaj rad predstavlja novi jezički resurs za pretraživanje i istraživanje verbalnih aspektnih parova u BCS (bosanskom, hrvatskom i srpskom), kreiran korišćenjem principa Lingvističkih Povezanih Otvorenih Podataka (LLOD). Pošto ne postoji resurs koji bi pomogao učenicima bosanskog, hrvatskog i srpskog kao stranih jezika da prepoznaju aspekt glagola ili njegove parove, kreirali smo novi resurs koji će korisnicima pružiti informacije o aspektu, kao i link ka aspektnim parovima glagola. Ovaj resurs takođe sadrži spoljne linkove ka monolingvalnim rečnicima, Wordnetu i BabelNetu. ...
    aspekt glagola, lingvistički otvoreni podaci, BCS

  • Serbian NER&Beyond: The Archaic and the Modern Intertwinned

    Branislava Šandrih Todorović, Cvetana Krstev, Ranka Stanković, Milica Ikonić Nešić (2021)
    U ovom radu predstavljamo srpski književni korpus koji se razvija pod okriljem COST Akcije „Distant Reading for European Literary History” CA16204. Koristeći ovaj korpus romana napisanih pre više od jednog veka, razvili smo i učinili javno dostupnim Sistem za prepoznavanje imenovanih entiteta (NER) obučen da prepozna 7 različitih tipova imenovanih entiteta, sa konvolucionom neuronskom mrežom (CNN), koja ima F1 rezultat od ≈91% na test skupu podataka. Ovaj model je dalje ocenjen na posebnom skupu podataka za evaluaciju. Završavamo poređenje ...
    ... ikonic.nesic@fil.bg.ac.rs Abstract In this work, we present a Serbian litera- ry corpus that is being developed under the umbrella of the “Distant Reading for European Literary History” COST Acti- on CA16204. Using this corpus of novels written more than a century ago, we ha- ve developed and made publicly ...
    ... open-source collection, named European Literary Text Col- lection (ELTeC), containing linguistically anno- tated sub-collections of 100 novels per language written more than 100 years ago. In this paper, we present a collection of Ser- bian texts in this corpus, named SrpELTeC. Alongside, we describe ...
    ... spaCy as a framework for developing a Serbian NER model on a collection old literary texts. 3 Serbian Collection in the ELTeC As described earlier in Section 1, the focus of the COST Action CA16204 is to compile the ELTeC corpus containing collections of old European novels published between 1840 and ...

  • A Lexical Approach to Acronyms and their Definitions

    Cvetana Krstev, Duško Vitas, Ranka Stanković (2015)
    In this paper we present a comprehensive approach to acronyms for Natural-Language Processing (NLP) of Serbian texts. The proposed procedure includes extraction of acronyms and their definitions that are usual Multi-Word Units (MWUs), shallow parsing of MWUs that enables MWU lemmatization and production of entries in morphological electronic dictionaries, both for MWU and acronyms, that are provided with grammatical, syntactic, semantic and domain information. This approach enables representation that reflects complex relations between acronyms and their definitions.
    ... +DOM=Mil +ACR=KFOR:ms3:ms7 3. Used Resources and Tools Corpus: As a corpus we have used an excerpt from the Corpus of Contemporary Serbian3 that has more than 22 million simple word forms (more than one million sen- tences). This corpus contains 70% of newspaper texts (57% daily, 8% weekly and ...
    ... 21 multiple entries 19 e-dict_MW_forms 9,956 6 corpus 23MW acr. inflects 333 acronyms 987 acr. in masc. 323 acr. in fem. 64 acr. both m&f 35 acr. in plural 2 7 acronyms 987 e-dict_acr_forms 6,946 Table 1: Results of the procedure on our corpus The majority of (acronym, definition) pairs represent ...
    ... logical dictionary of multi-word units. In IceTAL, vol- ume 6233 of LNCS. Springer. Krstev, C. and D. Vitas, 2005. Corpus and Lexicon – Mu- tual Incompletness. In Proc. of the Corpus Linguistics Conference, Birmingham. Liberman, Mark Y and Kenneth W Church, 1992. Text analysis and word pronunciation ...

  • Using Query Expansion for Cross-Lingual Mathematical Terminology Extraction

    Velislava Stoykova, Ranka Stanković (2018)

  • Using Lexical Resources for Irony and Sarcasm Classification

    Miljana Mladenović, Cvetana Krstev, Jelena Mitrović, Ranka Stanković (2017)
    The paper presents a language dependent model for classification of statements into ironic and non-ironic. The model uses various language resources: morphological dictionaries, sentiment lexicon, lexicon of markers and a WordNet based ontology. This approach uses various features: antonymous pairs obtained using the reasoning rules over the Serbian WordNet ontology (R), antonymous pairs in which one member has positive sentiment polarity (PPR), polarity of positive sentiment words (PSP), ordered sequence of sentiment tags (OSA), Part-of-Speech tags of words (POS) ...
    ... linguistic patterns were suggested for recognition of ironic statements in a corpus of tweets in Chinese, while authors in [32] used the pattern “about as * as *” as a query sent to the Google Search engine, in order to obtain a corpus of examples of the rhetorical figure of comparison (simile) which was later ...
    ... “figuratively” [14]. Authors in [29] used a microblogging platform called Plurk, similar to Twitter, to create a corpus of ironic statements. Other on- line resources can also be used for ironic corpus creation – Google Books search was used in the work by [21], where the phrase “said sarcastically” was used ...
    ... [4],[12],[28], where irony and sarcasm were treated in the same way in classification tasks. Similarly as in the case of building a corpus of ironic tweets, a corpus of sarcastic tweets has been generated based on online search with geolocation and time constraints, using the hashtag #sarkazam(sarcasm) ...

  • Претрага корпуса заснована на употреби екстерних лексичких ресурса путем веб-сервиса

    Милош Утвић, Ранка Станковић, Александра Томашевић, Михаило Шкорић, Биљана Лазић (2019)
    У раду се разматра хибридни приступ претрази корпуса, илустрован на примеру алатки OCWB и NoSketch Engine, примењених на специјални корпус из области рударства (РудКор) и Корпус савременог српског језика (СрпКор). Разматрани приступ комбинује постојеће могућности алатки OCWB и NoSketch Engine, које своју претрагу заснивају на лингвистичкој анотацији корпуса, са новим могућностима претраге у виду консултовања екстерних језичких ресурса (морфолошки електронски речници српског језика и лексичка база података Српски ворднет). Хибридни приступ је реализован надоградњом вебсучеља која поменуте алатке користе ...
    корпус, рударство, претраживање информација, проширивање упита, лексички ресурси, лексичке релације
    ... Lazić THE CORPUS SEARCH BASED ON USAGE OF EXTERNAL LEXICAL RESOURCES THROUGH WEB SERVICES Summary Тhis paper explores a hybrid approach to corpus search illustrated by application of corpus tools OCWB and NoSketch Engine to special corpus for mining domain, RudKor, as well as to Corpus of Contemporary ...
    ... Architecture for the New Millennium”, In: Proceedings of the Corpus Linguistics 2011 Conference, Birmingham, University of Birmingham. Еверт 2019: Stefan Evert and The OCWB Development Team, CQP Query Lan- guage Tutorial, The IMS Open Corpus Workbench (CWB 3.4.16), May 2019, http://cwb.sourceforge. ...
    ... i.org/10.18485/msc.2018.47.3.ch7. Харди 2012: Andrew Hardie, “CQPweb – Combining Power, Flexibility and Us- ability in a Corpus Analysis Tool”, International Journal of Corpus Linguis- tics. 17/3, 380–409. Шмид 1997: Helmut Schmid, „Probabilistic Part-of-Speech Tagging Using Deci- sion Trees”, In ...

  • Digital Library From A Domain Of Criminalistics As A Foundation For A Forensic Text Analysis

    Dalibor Vorkapić, Aleksandra Tomašević, Miljana Mladenović, Ranka Stanković, Nikola Vulović (2017)
    U ovom radu predstavljen je model koji omogućava prikupljanje, pripremu, opis metapodataka, upravljanje i eksploataciju, uključujući pretragu punog teksta dokumenata iz domena kriminalistike napisanih na srpskom jeziku. Predloženi pristup primenjuje se na veb portalu koji sakuplja različite tekstove nastale iz časopisa Akademije za kriminalistiku i policijske studije, Krivičnog zakona Srbije, konferencija „Tara“ i „Reiss“, kao i iz nekih doktorskih disertacija vezanih za ovu oblast istraživanje. Nakon obrade teksta, korpus koji sadrži preko 5500 stranica običnog teksta, kreiran je i ...
    Omeka, Wordnet, pretraga punog teksta, morfološka i semantička pretraga teksta, proširenje upita
    ... “reality” begin? BAAL Conference on The Ethics of Online Research Methods. Cardiff 9 Cvetana Krstev, Duško Vitas, “Corpus and Lexicon - Mutual Incompletness ”, in Proceedings of the Corpus Linguistics Conference, 14-17 July 2005, Birmingham, eds. Pernilla Danielsson and Martijn Wagenmakers, ISSN 1747-9398 ...
    ... Automata, Text and Electronic Dictionaries, Faculty of philology, Belgrade, 2008. 2. Cvetana Krstev, Duško Vitas, “Corpus and Lexicon - Mutual Incompletness ”, in Proceedings of the Corpus Linguistics Conference, 14-17 July 2005, Birmingham, eds. Pernilla Danielsson and Martijn Wagenmakers, ISSN 1747-9398 ...
    ... code of Serbia, the “Tara” and “Reiss” conferences, and from some of PhD dissertations related to this field of research. After text processing, a corpus containing over 5500 pages of plain text is created and prepared for publication as an online resource for full text search using Omeka, an open ...

  • Development of Open Educational Resources (OER) for Natural Language Processing

    Cvetana Krstev, Biljana Lazić, Ranka Stanković, Giovanni Schiuma, Miladin Kotorčević (2015)
    In this paper we present the development of an online course at the edX BAEKTEL platform named “Lexical Recognition in the Natural Language Processing (NLP)”. It is based on the course of the same name for PhD studies at the University of Belgrade, Faculty of Philology. There are not many courses in Computational Linguistics (CL) on OER platforms, and there is none in Serbian either for CL or NLP. We have developed this course in order to improve this ...
    E-Learning, Open Educational Resources, Computational Linguistics, Lexical Resources, edX
    ... improve this situation as it can prove useful both for linguists working in corpus linguistics and computer scientists developing NLP applications. The participant will become familiar with the use of Unitex, the corpus processing system for which many valuable resources for Serbian were already ...
    ... oriented processing of texts in human languages. As an illustration, a string oriented web interface to the Corpus of Contemporary Serbian 13 is presented. [13][14] 2. Unitex corpus processing system is presented from the practical point of view: how to install it and start working with it ...
    ... within BAEKTEL project of its OER version within the edX BAEKTEL platform. 2 The main features of Unitex, an open access and open source corpus processing system, are presented in Section 4. Section 5 presents course content with didactic criteria and specific formats used in the OER course ...

  • Old or New, We Repair, Adjust and Alter (Texts)

    Cvetana Krstev, Ranka Stanković (2020)
    U ovom radu predstavljamo kako se e-rečnici i kaskade transduktora konačnih stanja implementirani u alatu Unitex mogu koristiti za rešavanje tri problema transformacije teksta: ispravljanje tekstova nakon OCR-a, vraćanje dijakritičkih znakova i prebacivanje između različitih jezičkih varijanti.
    ispravka teksta, OCR greške, restauracija dijakritika , jezičke varijante, elektronski rečnik, transduktori konačnih stanja
    ... Serbian morphological electronic dictionaries (SMD) (Krstev, 2008); 2 This corpus is a part of the European Literary Text Collection corpus (ElTEC) developed in the scope of the COST action 16204 Distant Reading for European Literary History (d-reading). 64 Infotheca Vol. 19, No. 2, December 2019 Scientific ...
    ... present the solution for detecting and correcting OCR errors developed for the compilation of the corpus of Serbian novels written and published in the period 1840–1920.2 The novels selected for this corpus were mainly printed in Cyrillic script (only a few of them were in Latin script). When scanning ...
    ... performed by finite-state transducers (FST) implemented in Unitex.3 A separate FST is written for each replacement 3 UnitexGramLab, a lexicon-based corpus processing suite. Infotheca Vol. 19, No. 2, December 2019 65 Krstev C., Stanković R., “Old or new, we repair . . . ”, pp. 61–80 a non-valid ...

  • Classification of Terms on a Positive-Negative Feelings Polarity Scale Based on Emoticons

    Mihailo Škorić (2017)
    The goal of this paper is to draw attention to the possibility of using emoticon-riddled text on the web in language-neutral sentiment analysis. It introduces several innovations in the existing framework of research and tests their effectiveness. It also presents a software tool especially made for that purpose, explains how it builds a database with sentimental value of terms and offers the user manual. Finally, it presents a software tool that tests the new database and gives some examples ...
    data mining, information extraction, emotions, text on the web
    ... databases for this study were created, from the collection of the corpus to the export of completed database, which can then be used in several ways. 2.1 Collecting textual corpus The basic idea was for the database to be based on a corpus of texts containing determiners which express positive or negative ...
    ... pendent, the system would be language-independent as well. If it turns out to be valid, this method could allow machine learning the usage of huge corpus of texts that are pre-labeled with determiners. 1.1 Review of their former similar studies In 2005 a series of experiments with the classification ...
    ... and language-neutral determiner strings will be used. Goal is to create a fully language-independent system that would greatly broaden the possible corpus. 1 Users of Twitter platform have an option to additionally mark their posts with tags so that posts that talk about a certain topic can be found ...

  • Using English Baits to Catch Serbian Multi-Word Terminology

    Cvetana Krstev, Branislava Šandrih, Ranka Stanković (2018)
    In this paper we present the first results in bilingual terminology extraction. The hypothesis of our approach is that if for a source language domain terminology exists as well as a domain aligned corpus for a source and a target language, then it is possible to extract the terminology for a target language. Our approach relies on several resources and tools: aligned domain texts, domain terminology for a source language, a terminology extractor for a target language, and a ...
    aligned texts, word alignment, terminology extraction, electronic dictionaries, morphological inflection
    ... overall design of our system (Fig- ure1) is as follows: 1. Input: • A sentence-aligned domain-specific corpus in- volving a source and a target language. We will denote an entry in this corpus with S(text.align) ↔ T (text.align); • A list of terms from the same domain in a source language (both s ...
    ... extracted from the target part of the aligned corpus having some expected syntac- tic structure. We will denote an entry from this list with T (term.extract). 2. Processing: • Aligning bilingual chunks (possible translation equivalents) from the aligned corpus. We will denote aligned chunks with S(align ...
    ... contained 491,990 translation pair candidates. We decided to enrich corpus with additional parallel lists (described in Subsection 4.4.) since we observed certain improvement in evaluations of translation quality. First we splitted corpus of aligned sentences into three disjoint parts: training (80%), ...