Automated identification of borrowings in multilingual wordlists

Abstract

Although lexical borrowing is an important aspect of language evolution, there have been few attempts to automate the identification of borrowings in lexical datasets. Moreover, none of the solutions which have been proposed so far identify borrowings across multiple languages. This study proposes a new method for the task and tests it on a newly compiled large comparative dataset of 48 South-East Asian languages. The method yields very promising results, while it is conceptually straightforward and easy to apply. This makes the approach a perfect candidate for computer-assisted exploratory studies on lexical borrowing in contact areas.

Lexibank: A public repository of standardized wordlists with computed phonological and lexical features

“…Given that computational methods for the detection of cognates are still not able to compete with experts (List et al 2017), our collection thus offers rich material to test and train new methods in the future. In a similar way -given that the collection uni es data on a global basis -scholars can use the data collection to test new methods for the automated identi cation of borrowings (Zhang et al 2019;List and Forkel 2021b), or to expand upon previous approaches to the automated detection of contact areas (Gast and Koptjevskaja-Tamm 2018;Matsumae et al 2021;Ranacher et al 2021). In addition, we illustrate how the data can be used to automatically extract various phonological and lexical features for individual language varieties.…”

“…Lexibank and lexical data in CLDF formats have been promoted in several ways so far. First, we have conducted detailed studies in which CLDF formats are used along with CLDFBench and the pylexibank software package, illustrating how data aggregation can be successfully carried out (List et al 2018;Rzymski et al 2020), or showing how data can be supplemented in transparent CLDF formats (Wu et al 2020;List and Forkel 2021b). Second, we have created certain agship projects which showcase speci c aspects of CLDF and the advantage of using integrated data (Geisler et al 2020;Ferraz Gerardi et al 2021).…”

Lexibank, a public repository of standardized wordlists with computed phonological and lexical features

“…There are discrete features on sound inventory sizes (1-7, number of vowels, consonants, etc. ), there are various features on special sound types or individual specific sounds (8)(9)(10)(11)(12)(13)(14)(15)(16)(17)(18)(19), there are three prosodic features (20)(21)(22), and eight features pertaining to specific sound-meaning relations (also termed "sound symbolism", [23][24][25][26][27][28][29][30].…”

“…Lexibank and lexical data in CLDF formats have been promoted in several ways. First, we have conducted detailed studies in which CLDF formats are used along with CLDFBench and the PyLexibank software package, illustrating how data aggregation can be successfully carried out 60,61 , or showing how data can be supplemented in transparent CLDF formats 21,68 Second, we have created certain flagship projects which showcase specific aspects of CLDF and the advantage of using integrated data 99,100 . Third, we have conducted projects with students and young scholars, who were trained to use our new resources and encouraged to share their knowledge in the form of small blog posts (published at https://calc.hypotheses.org) along with new datasets which bachelor, doctoral, and master students lifted themselves assisted by our team 70,[101][102][103] .…”

“…Given that computational methods for the detection of cognates are still not able to compete with experts 19 , our collection thus offers rich material to test and train new methods in the future. Similarly -given that the Lexibank collection unifies data on a global basis -scholars can use the data collection to test new methods for the automated identification of borrowings 20,21 , or to expand upon previous approaches to the automated detection of contact areas [22][23][24] . In addition, we illustrate how the data can be used to automatically extract various phonological and lexical features for individual language varieties.…”