Corpus-based comparison of contemporary Croatian, Serbian and Bosnian

This paper explores the differences between three Slavic languages: Bosnian, Croatian and Serbian, drawing on the Southeast European Times newspaper corpus, translated to each language from the source English text and consisting of approximately 330,000 tokens for each language. The paper is an effort intended to contribute to the establishment of the criteria and methodology for measuring similarities between these languages. The differences were explored at five levels: at the level of phonology, morphology, lexis, syntax and semantics. Empirical analysis has shown that a huge portion of differences across the three languages are systematic and regular, and as such, could be formalized for automatic translation/generation. The results of this study and of similar future corpus-based studies can be used in developing NLP tools such as annotating tools, e-dictionaries, text summarizers, machine translation systems, computerassisted language learning etc. for the three languages, as well as further linguistic investigation of their mutual relationship.