مدلسازي آماري زبان فارسي به كمك دسته بندي منطقي دستوري براي بازشناسي گفتار پيوسته

This paper investigates a category-based statistical language model for Persian continuous speech recognition. The language models are based on variable-length category-based n-grams. Instead of finding patterns among individual words, a language model may be designed to discover relationships between word categories. In this research, this has been accomplished in 3 steps; clustering words into groups, exploitation of the statistical language model employing tree data structure and its application to the recognition system. The most important advantage of this model has found to be its ability to return to correct path in situations where ordinary statistical modeling does not allow it.