A Large-Scale Study of Machine Translation in Turkic Languages

Abstract

Recent advances in neural machine translation (NMT) have pushed the quality of machine translation systems to the point where they are becoming widely adopted for building competitive systems. However, there is still a large number of languages that are yet to reap the benefits of NMT. In this paper, we provide the first large-scale case study of the practical application of MT in the Turkic language family in order to realize the gains of NMT for Turkic languages under high-resource to extremely low-resource scenarios. In addition to presenting an extensive analysis that identifies the bottlenecks towards building competitive systems to ameliorate data scarcity, our study has several key contributions, including, i) a large parallel corpus covering 22 Turkic languages consisting of common public datasets in combination with new datasets of approximately 2 million parallel sentences, ii) bilingual baselines for 26 language pairs, iii) novel high-quality test sets in three different translation domains and iv) human evaluation scores. All of our data, software and models are publicly available. 1

The Fourth Workshop on Insights from Negative Results in NLP

“…For all languages except Uzbek, we use the WMT shared task data from He et al (2020). For Uzbek, we use the Turkic Interlingua corpus (Mirzakhalov et al, 2021).…”

What a Creole Wants, What a Creole Needs

“…The field of natural language processing (NLP) has become aware that most of the world's languages are unfortunately under-represented, or entirely absent, from the field's body of work (Joshi et al, 2020). In recent years, there has been a push in efforts to ameliorate this discrepancy (Nekoto et al, 2020;Mirzakhalov et al, 2021;Ogueji et al, 2021). Among these low-resourced languages 1 are Creole languages, which are particularly under-resourced due to barriers like societal stigma (Siegel, 1999), despite the fact that these languages are spoken by many people globally.…”