Research on Confusion Network Algorithm Based on Minimum Bayes Risk Decision Rule

摘要

In mandarin large vocabulary continuous speech recognition, we can obtain recognition result which word error rate(WER) is minimum by using minimum bayes risk(MBR) decoding strategy. One method of MBR decoding is that the word lattice can be transformed into confusion network in order to obtain the recognition result with minimum WER. According to the characteristic of mandarin, we proposed an improved confusion network generation algorithm based on prevenient works. The experimental results of mandarin large vocabulary continuous speech recognition show that the improved algorithm yields a lower WER than the MAP recognition and prevenient two confusion network generation algorithms.