The Nunavut Hansard Inuktitut–English Parallel Corpus 3.0 with Preliminary Machine Translation Results

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 2562–2572 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 2562 The Nunavut Hansard Inuktitut–English Parallel Corpus 3.0 with Preliminary Machine Translation Results Eric Joanis, Rebecca Knowles, Roland Kuhn Samuel Larkin, Patrick Littell, Chi-kiu Lo, Darlene Stewart National Research Council Canada 1200 Montreal Road, Ottawa ON, K1A 0R6 firstname.lastname@nrc-cnrc.gc.ca, chikiu.lo@nrc-cnrc.gc.ca Jeffrey Micher US Army Research Laboratory 2800 Powder Mill Road, Adelphi MD 20783 jeffrey.c.micher.civ@mail.mil Abstract The Inuktitut language, a member of the Inuit-Yupik-Unangan language family, is spoken across Arctic Canada and noted for its morphological complexity. It is an official language of two territories, Nunavut and the Northwest Territories, and has recognition in additional regions. This paper describes a newly released sentence-aligned Inuktitut–English corpus based on the proceedings of the Legislative Assembly of Nunavut, covering sessions from April 1999 to June 2017. With approximately 1.3 million aligned sentence pairs, this is, to our knowledge, the largest parallel corpus of a polysynthetic language or an Indigenous language of the Americas released to date. The paper describes the alignment methodology used, the evaluation of the alignments, and preliminary experiments on statistical and neural machine translation (SMT and NMT) between Inuktitut and English, in both directions. Keywords: Inuktitut, SMT, NMT, sentence alignment, machine translation for polysynthetic languages, Indigenous languages 1. Introduction This paper describes the release of version 3.0 of the Nunavut Hansard Inuktitut–English Parallel Corpus, a sentence-aligned collection of the proceedings of the Leg- islative Assembly of Nunavut (Nunavut Maligaliurvia).1 The Nunavut Hansard 3.0 corpus consists of 8,068,977 Inuktitut words and 17,330,271 English words and covers the proceedings of the 687 debate days from April 1, 1999 to June 8, 2017. We include an account of several automatic sentence align- ment techniques examined, along with a summary of exper- iments in which gold standard sentence alignments created by Inuktitut–English bilinguals were compared with those created by our chosen automatic alignment technique. We also present baseline statistical and neural machine trans- lation experiments on the corpus, showing the positive ef- fects of improved alignment and corpus size increase, and provide a brief discussion of challenges of machine transla- tion into morphologically complex languages. This corpus release marks a move away from machine translation “low- resource” status for this language pair and domain. 2. Background 2.1. Nunavut and Inuktut Nunavut is a territory of Canada that comprises most of the Canadian Arctic archipelago. Separated from the North- west Territories on April 1, 1999, it is the largest of Canada’s territories. According to the 2016 census, 65.3% of the population of Nunavut (of 35,944) had a variety of Inuktut 1The corpus is available at the NRC Digital Repository: https://doi.org/10.4224/40001819 (CC-BY-4.0 license). (Inuktitut or Inuinnaqtun) as their first language, and 89.0% of Inuit in Nunavut could speak Inuktut at a conversational level or above (Lepage et al., 2019). Figure 1: The Nunavut territory and its location within Canada. Map data ©2019 Google, INEGI Inuktut (not to be confused with …

Content truncated to 500 words