CogNet: A Large-Scale Cognate Database

Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 3136–3145 Florence, Italy, July 28 - August 2, 2019. c⃝2019 Association for Computational Linguistics 3136 CogNet: a Large-Scale Cognate Database Khuyagbaatar Batsuren† Gábor Bella† Fausto Giunchiglia†§ DISI, University of Trento, Trento, Italy† Jilin University, Changchun, China§ {k.batsuren; gabor.bella; fausto.giunchiglia}@unitn.it Abstract This paper introduces CogNet, a new, large-scale lexical database that provides cognates—words of common origin and meaning—across languages. The database currently contains 3.1 million cognate pairs across 338 languages using 35 writing sys- tems. The paper also describes the automated method by which cognates were computed from publicly available wordnets, with an accuracy evaluated to 94%. Finally, statistics and early insights about the cognate data are presented, hinting at a possible future exploitation of the resource1 by various fields of lingustics. 1 Introduction Cognates are words in different languages that share a common origin and the same meaning, such as the English letter and the French lettre. Cognates and the problem of cognate identifica- tion have been extensively studied in the fields of language typology and historical linguistics, as cognates are considered useful for research- ing the relatedness of languages (Bhattacharya et al., 2018). Cognates are also used in computa- tional linguistics, e.g., for lexicon extension (Wu and Yarowsky, 2018) or to improve cross-lingual NLP tasks such as machine translation or bilingual word recognition (Kondrak et al., 2003; Tsvetkov and Dyer, 2015). Despite the interest in using cognate data for research, state-of-the-art cognate databases have had limited practical uses from an applied perspec- tive, for two reasons. Firstly, popular cognate- coded databases that are used in historical lin- guistics, such as ASJP (Wichmann et al., 2010), 1The CogNet resource and WikTra tool are available on http://cognet.ukc.disi.unitn.it. IELex2, or ABVD (Greenhill et al., 2008), cover only the small set of 225 Swadesh basic concepts, although with an extremely wide coverage of up to 4000 languages. Secondly, in these databases, lex- ical entries that belong to scripts other than Latin or Cyrillic mostly appear in phonetic transcription instead of using their actual orthographies in their original scripts. These limitations prevent such re- sources from being used in real-world computa- tional tasks on written language. This paper describes CogNet, a new large-scale, high-precision, multilingual cognate database, as well as the method used to build it. Our main technical contributions are (1) a general method to detect cognates from multilingual lexical re- sources, with precision and recall parametrable ac- cording to usage needs; (2) a large-scale cognate database containing 3.1 million word pairs across 338 languages, generated with the method above; (3) WikTra, a multilingual transliteration dictio- nary and library derived from Wiktionary data; and (4) an online platform that lets users explore the resource. The paper is organised as follows. Section 2 presents the state of the art. Section 3 describes the main cognate discovery algorithm and sec- tion 4 the way various forms of evidence used by the algorithm are computed. The method is parametrised and the results are …

Content truncated to 500 words