Revitalization of Indigenous Languages through Pre-processing and Neural Machine Translation: The case of Inuktitut

Proceedings of the 28th International Conference on Computational Linguistics, pages 4661–4666 Barcelona, Spain (Online), December 8-13, 2020 4661 Revitalization of Indigenous Languages through Pre-processing and Neural Machine Translation: The case of Inuktitut Ngoc Tan Le UQAM / Montreal, Quebec, Canada le.ngoc tan@uqam.ca Fatiha Sadat UQAM / Montreal, Quebec, Canada sadat.fatiha@uqam.ca Abstract Indigenous languages have been very challenging when dealing with NLP tasks and applica- tions because of multiple reasons. These languages, in linguistic typology, are polysynthetic and highly inflected with rich morphophonemics and variable dialectal-dependent spellings; which affected studies on any NLP task in the recent years. Moreover, Indigenous languages have been considered as low-resource and/or endangered; which poses a great challenge for research re- lated to Artificial Intelligence and its fields, such as NLP and machine learning. In this paper, we propose a study on the Inuktitut language through pre-processing and neural machine translation, in order to revitalize the language which belongs to the Inuit family, a type of polysynthetic lan- guages spoken in Northern Canada. Our focus is concentrated on: (1) the preprocessing phase, and (2) applications on specific NLP tasks such as morphological analysis and neural machine translation, both for Indigenous languages of Canada. Our evaluations in the context of low- resource Inuktitut-English Neural Machine Translation, showed significant improvements of the proposed approach compared to the state-of-the-art. 1 Introduction North America and Canada represent a highly complex linguistic regions, with numerous languages and great linguistic diversity. Indigenous languages are spoken widely and are official languages in Nunavut and the Northwest Territories. Indigenous peoples are making efforts to revitalize and sustain their languages, although they face many difficulties. The case of Inuktitut language in Canada has recently been a good example to examine. This language belongs to the Inuktut - the term Inuktut is inclusive of all dialects used in Nunavut - one of the official languages of Nunavut, the largest territory of Canada, which can be written in syllabics or in roman orthography, and regional variations use different special characters and spelling conventions (Statistics-Canada, 2017). As Inuktitut is an official language in Nunavut, there exists some resources that are available in this language at a much larger scale than most other languages in the same family, notably a parallel corpus with English. The researchers have shown more interest in Inuktut languages in NLP communities to respond to an increasing demand for Indigenous language educational materials and technology. Previous studies have demonstrated that the development of Indigenous language technology faces many challenges on one hand, due to a high morpheme per word rate and the rich morphophonemics and variable dialectal-dependent spellings, a complex polysynthesis in linguistic typology, and on the other hand, due to the lack of orthographic standardization and a large dialect variation. Digital text and voice data are limited. It poses many greater challenges for NLP to develop applications for all users (Littell et al., 2018; Mager et al., 2018). This research paper examines the case of Inuktitut through experiments on the crucial phase as pre- …

Content truncated to 500 words