We describe methods for estimation the phonemic, words and lexical ambiguity for automatic speech recognition systems. This methods are use the phonemic structure of spoken words, size of dialogue vocabulary, probability ...