Detecting Lexical Borrowings from Dominant Languages in Multilingual Wordlists

Proceedings of the 17th Conference of the European Chapter of the Association for Computational Linguistics, pages 2599–2605 May 2-6, 2023 ©2023 Association for Computational Linguistics Detecting Lexical Borrowings from Dominant Languages in Multilingual Wordlists John E. Miller Artificial Intelligence/Engr. Pontificia Universidad Católica del Perú San Miguel, Lima, Peru jemiller@pucp.edu.pe Johann-Mattis List Chair of Multilingual Computational Linguistics University of Passau Germany mattis.list@uni-passau.de Abstract Language contact is a pervasive phenomenon reflected in the borrowing of words from donor to recipient languages. Most computational approaches to borrowing detection treat all lan- guages under study as equally important, even though dominant languages have a stronger im- pact on heritage languages than vice versa. We test new methods for lexical borrowing detec- tion in contact situations where dominant lan- guages play an important role, applying two classical sequence comparison methods and one machine learning method to a sample of seven Latin American languages which have all borrowed extensively from Spanish. All methods perform well, with the supervised ma- chine learning system outperforming the classi- cal systems. A review of detection errors shows that borrowing detection could be substantially improved by taking into account donor words with divergent meanings from recipient words. 1 Introduction Language contact is one of the most pervasive lin- guistic phenomena. It is the first factor that needs to be excluded when searching for genetic rela- tions among the languages of the world, and it needs to be controlled for when searching for cross- linguistic universals. Any study trying to explain how humans use language must take language con- tact into account. It is also a unique witness of ancient contacts in human prehistory. Language contact is most prominently reflected in lexical borrowing, the transfer of words from a donor language to a recipient language. Although research on the computational handling of lexical borrowing has made some progress of late, com- putational approaches to the investigation of lan- guage contact are still in their infancy (List, 2019). Specifically methods that could infer the direction of borrowings have not been proposed so far. While numerous case studies investigate the influence of dominant languages on heritage languages (Proc- hazka and Vogl, 2017; Meisel, 2018), we lack au- tomated methods that could be used to study the influence of particular dominant languages in lin- guistically diverse areas of the world. The number of standardized cross-linguistic wordlist collections has greatly increased over the past years (Rzymski et al., 2020; List et al., 2022a), with standard formats (Forkel et al., 2018) accom- panied by software tools that allow scholars to pre- pare and curate standardized wordlists in an effi- cient manner (Forkel and List, 2020). It would be useful to have a computer assisted tool for linguists to detect which words in a cross-linguistic region have been borrowed from a dominant language and with further advances, a tool for inclusion in language contact assessment or in computational cladistics workflows. Here, we compare the suitability of three differ- ent systems to address this task – two systems based …

Content truncated to 500 words