The article contains the analysis of existing methods of machine transliteration. General structure of cross-lingual transliteration system is represented. The author considers statistic systems as well as rule-based systems, and explores two main classes of statistical methods of transliteration learning: discriminative and generative methods.
