Parallel corpus of abstracts on management Kormana 1.0

  Avtorji

Paradiž, Martina ; et al.prikaži vse

Paradiž, Martina ; Melanšek, Aleksandra ; Romih, Miro ; Grilc, Iztok ; Samoiliuk, Viktoriia ; Erjavec, Tomaž

  Identifikator vnosa

http://hdl.handle.net/11356/2643

 Datum objave

2026-11-14

 Vrsta

corpus, text

 Velikost

1822 texts, 19566 translationUntis, 843085 tokens

 Jezik(i)

English , Slovenian

 Opis

The Kormana 1.0 corpus contains 1,822 English-Slovene and Slovene-English translation pairs of article abstracts from 12 open-access academic journals in the field of management. Of the 1,822 pairs, 1,098 are English originals with Slovene translations and 724 are Slovene originals with English translations. The corpus contains about 843,000 tokens. The corpus can be useful for research into terminology, phraseology, machine translation, translation studies and translator training. The texts are accompanied by metadata on the title, author, year of publication, name of journal, and translation direction. The texts were extracted from PDF files and may therefore contain character noise. The Slovene and English parts of the corpus are sentence-aligned, with the alignment manually checked. The linguistic annotation was perfomed by CLASSLA version 2.2.3 (https://github.com/clarinsi/classla) for Slovenian and by STANZA version 1.15.0 (https://stanfordnlp.github.io/stanza/) for English. The annotations comprise tokenisation, sentence segmentation, lemmatisation, the language-specific PoS tag (XPOS), Universal Dependencies PoS and morphological features and dependency syntax, as well as Named Entities (NE). XPOS for Slovenian are the MULTEXT-East MSDs and Penn Treebank PoS tags for English. As CLASSLA and STANZA have slighlty different formats, the encoding of the English CoNLL-U file was modified to be compatible with the Slovenian one. The detailed English NE categories were also reduced to the standard four-class NEs (PER, ORG, LOC, MISC) as used by CLASSLA. The corpus is available in CoNLL-U format and as vertical files (derived from CoNLL-U) for use in CWB-tye concordancers, separately by language. Acknowledgements: We would like to thank the students of the master’s degree course in Linguistic Mediation and Translation at the University of Primorska for their contribution to data collection: Monika Briško, Sašo Arnejčič, Demi Milovanović, Marija Shoposka, Kaja Černigoj.

 Izdajatelj

Amebis, d. o. o., Kamnik

 

University of Primorska, Faculty of Humanities

 Zahvala

Jožef Stefan Institute CLARIN "CLARIN.SI"

 Ključne besede

parallel corpus abstracts management translation domain-specific corpus

 Zbirke

CLARIN.SI data & tools

Prikaži polni zapis vnosa