Analysis of long-distance word dependencies and pronunciation variability at conversational Russian speech recognition

The key issues of conversational Russian speech processing at phonemic and language model levels are considered in the work. The multiple transcriptions for modeling word pronunciation variety and joint application of statistic and syntactic analysis of training text data for modeling long-distance grammatical relations between words in the phrase are proposed. The word error rate of the developed speech recognition system was 33% for the collected conversational speech corpus.