Finite-state morphological analysis for Gagauz

Finite-state morphological analysis for Gagauz Sevilay Bayatli,⋆ Güllü Karanfil,† Memduh Gökırmak,‡ Francis M. Tyers♣ ⋆ Faculty of Electrical and Computer Engineering, School of Science and Engineering, Altınbaş University † Anadolu BİL Meslek Yüksekokulu, Uygulamalı Rusça ve Çevirmenlik, İstanbul Aydin University ‡ Department of Computer Engineering, School of Engineering, İstanbul Teknik Üniversitesi ♣ School of Linguistics, Faculty of the Humanities, Higher School of Economics sevilaybayatli@gmail.com, gullukaranfil@hotmail.com, memduhg@gmail.com, ftyers@hse.ru Abstract This paper describes a finite-state approach to morphological analysis and generation of Gagauz, a Turkic language spoken in the Republic of Moldova. Finite-state approaches are commonly used in morphological modelling, but one of the novelties of our approach is that we explicitly handle orthographic errors and variance, in addition to loan words. The resulting model has a reasonable coverage (above 90%) over a range of freely-available corpora. Keywords: morphology, Gagauz, finite-state transducer 1. Introduction Gagauz is a Turkic language of the Oghuz group spoken by approximately 140,920 worldwide and 105,000 in the Re- public of Moldova where it is co-official with Moldovan1 in the autonomous region of Gagauzia. Most speakers are bilingual in Russian and education takes place almost en- tirely in Russian. This leads to a number of issues with at- tempting to create a wide-coverage morphological model, as text in the wild is often less orthographically accurate than in languages which are widely taught. There is a fur- ther issue that keyboards may not be easily available, and so the choice might be between a Turkish keyboard which does not include ţ and a Moldovan keyboard which does not include ı, ö, ü. The paper is laid out as follows, in Section 2. we describe give a brief typological description of Gagauz, in Section 3., in Section 4. we describe the evaluation and results then in Sections 5. and 6. we give some thoughts on further work and conclusions. 2. Gagauz To our knowledge to date there has been no computational linguistic work on Gagauz. There are two available gram- mars, one in Russian (Покровская, 1964) and one in Turk- ish (Özkan, 1996). The former describes the literary lan- guage, where the latter takes a comparative approach, com- paring Gagauz with Turkish. There have also been a number of works on Gagauz lexicography, for example Каранфил (2009). As with other Turkic languages, Gagauz is an SOV lan- guage. It has no gender, plural marking and seven cases (nominative, genitive, dative, accusative, locative, ablative and instrumental). Vowel harmony is applied in affixes. In- flectional marking is used to indicate possession and also for agreement within subordinate clauses, işledini gördüm ‘I saw that you were working’ (lit. ‘I saw your working’). 1Moldovan is one of the official names for Romanian in the Re- public of Moldova. The discussion of if Romanian and Moldovan are separate languages is a sociopolitical one and so for the pur- poses of this article we consider the names to be equivalent. Figure 1: Location of the Gagauz speaking area (gag) within the Black Sea region, relative to …

Content truncated to 500 words