Apparatus for correcting error in speech recognition

摘要

An apparatus for correcting errors in speech recognition is provided. The apparatus includes a feature vector extracting unit extracting feature vectors from a received speech. A speech recognizing unit recognizes the received speech as a word sequence on the basis of the extracted feature vectors. A phoneme weighted finite state transducer (WFST)-based converting unit converts the recognized word sequence recognized by the speech recognizing unit into a phoneme WFST. A speech recognition error correcting unit corrects errors in the converted phoneme WFST. The speech recognition error correcting unit includes a WFST synthesizing unit modeling a phoneme WFST transferred from the phoneme WFST-based converting unit as pronunciation variation on the basis of a Kullback-Leibler (KL) distance matrix.

主权项

1. An apparatus for correcting errors in speech recognition, the apparatus comprising:
a feature vector extracting unit extracting feature vectors from a received speech; a speech recognizing unit recognizing the received speech as a word sequence on the basis of the extracted feature vectors; a phoneme weighted finite state transducer (WFST)-based converting unit converting the recognized word sequence recognized by the speech recognizing unit into a phoneme WFST; and a speech recognition error correcting unit correcting errors in the converted phoneme WFST, wherein the speech recognition error correcting unit comprises a WFST synthesizing unit modeling a phoneme WFST transferred from the phoneme WFST-based converting unit as pronunciation variation on the basis of a Kullback-Leibler (KL) distance matrix, and wherein the KL distance is calculated by an equation 1,KL⁡(L-Ci+R||L-Cj+R)=∑k=1M⁢ωi,k⁢KL⁡(bi,k||bj,k)+∑k=1M⁢ωi,k⁢log⁢⁢ωi,kωj,k.(1)