Estimation of language complexity for the human computer interection systems

Abstract

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 of words occurrences of spoken language and allows to predict the accuracy of word recognition and direction to increasing the accuracy on the phase of building the system. Speech recognition system is considering as the informational channel with the noise for transferring information and we estimate the information lost in this channel. The lost information can be a measure for estimation the uncertainty or complexity of language recognition for dialog systems. These methods make it possible to use linguistic knowledge and phonetic structure of words to improve the reliability of speech man-computer dialogue.