Computational Historical Linguistics and Language Diversity in South Asia

Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics Volume 1: Long Papers, pages 1396 - 1409 May 22-27, 2022 c⃝2022 Association for Computational Linguistics Computational historical linguistics and language diversity in South Asia Aryaman Arora Georgetown University aa2190@georgetown.edu Adam Farris San Mateo High School adamfarris@gmail.com Samopriya Basu University of North Carolina – Chapel Hill sampr0b@live.unc.edu Suresh Kolichala Microsoft suresh.kolichala@gmail.com Abstract South Asia is home to a plethora of languages, many of which severely lack access to new language technologies. This linguistic diver- sity also results in a research environment con- ducive to the study of comparative, contact, and historical linguistics—fields which necessi- tate the gathering of extensive data from many languages. We claim that data scatteredness (rather than scarcity) is the primary obstacle in the development of South Asian language tech- nology, and suggest that the study of language history is uniquely aligned with surmounting this obstacle. We review recent developments in and at the intersection of South Asian NLP and historical–comparative linguistics, describ- ing our and others’ current efforts in this area. We also offer new strategies towards breaking the data barrier. 1 Introduction South Asia1 is home to one-quarter of the world’s population and boasts immense linguistic diver- sity (Saxena and Borin, 2008; Bashir, 2016). With members of at least four top-level major linguistic families2 and several putative linguistic isolates, this region is a fascinating arena for linguistic re- search. The languages of South Asia, moreover, have a long recorded history, and have undergone complex change through genetic descent, sociolin- guistic interactions, and contact influence. Nevertheless, South Asian languages for the most part remain severely underdocumented (van Driem, 2008), and several languages with even official administrative status (e.g. Sindhi) are low- resource (if not data-scarce) for the purposes of all natural language processing tasks (Joshi et al., 1Roughly the Indian Subcontinent, or the geographic and cultural region enclosed by the Himalayas, the Indian Ocean, and the Hindu Kush. 2Indo-European (Indo-Aryan, Iranic, Nuristani), Dravid- ian, Austroasiatic (Munda, Khasian), Sino-Tibetan (several branches). Figure 1: The major language families of South Asia (Kolichala, 2016). 2020). This data scatteredness persists despite long native traditions of linguistic description, contin- ued language vitality with active use on the inter- net, and vast numbers of speakers (Rahman, 2008; Groff, 2017). We argue that the most basic problem in NLP/CL work on South Asian languages is not data scarcity, but data scatteredness. There is much data to be extracted for even the most endangered languages (e.g., Burushaski, a language isolate of the north- west), from annotated corpora and grammatical descriptions compiled by linguists, if only one is willing to wrangle idiosyncratic data formats and digitise existing texts. Thus far, commercial inter- ests and scientific agencies have only intermittently supported the development of language technology for the region—taking a new approach, we pro- pose a research programme from the perspective of computational historical linguistics, outlining current data gathering initiatives in this discipline and potential benefits to other work across NLP. 1396 Level Languages …

Content truncated to 500 words