Аңдатпа
We compare and discuss various approaches to the problem of part of speech (POS) tagging of texts written in Kazakh, an agglutinative and highly inflectional Turkic language. In Kazakh a single root may produce hundreds of word forms, and it is difficult, if at all possible, to label enough training data to account for a vast set of all possible word forms in the language. Thus, current state of the art statistical POS taggers may not be as effective for Kazakh as for morphologically less complex languages, e.g. English. Also the choice of a POS tag set may influence the informativeness and the accuracy of tagging.