Single state transducer model for Kazakh and Russian morphology

Abstract

This paper provides a broad overview of issues related to the construction of finite state transducers with one state for the two-level morphology of inflectional languages, particularly, the direct transformation of word endings to the grammatical characteristics. This problem has been studied on the base of the Kazakh and Russian languages, which are usually named the inflectional languages. The solution of this problem is the trivial Mealy automaton with one state, i.e. a single state transducer, and a multi-valued mapping method is used as well. We study the problem of completeness of the finite state transducers input for the analyzed languages. The determination of transducer input completeness for morphological analysis gives a guarantee that all the words of the analyzed language will be accepted. The problem of determining the completeness of the set of possible endings for agglutinative languages is a complex issue. In this article, we define the completeness of a set of endings in Kazakh language. The proposed technology is implemented for the Russian-Kazakh machine translation, a translation quality assessment performed by the method of BLEU.