Inflection of Czech Words Using POS-based n-gram

There are many more forms of words in Slavic languages com-pared to Germanic. The Czech language contains about 300K words in the basic form, and c. 5M words in various forms. The main idea of this paper consists in common n-gram lan-guage modelling based on several types (260) of classes based on parts of speech- POS. These depend on given case, gender, number, or voice. This n-gram is used for matching the best word sequence from a lattice generated by uniform-distributed word based speech recognizer. Index Terms: speech recognition, Slavic languages, parts of speec