Title :
An N-gram based Chinese syllable evaluation approach for speech recognition error detection
Author :
Wang, Xingjian ; Li, Lei
Author_Institution :
Center for Intell. Sci. & Technol., Beijing Univ. of Posts & Telecommun., Beijing, China
Abstract :
In order to find errors and correct words after Chinese speech recognition to improve its accuracy rate, an N-gram based phonetic syllable evaluation approach is proposed according to the conjunction rules in Chinese syllables. In this paper, Bigram and Trigram model are used with two kinds of smoothing methods for comparison. Evaluation results show that it can achieve a precision rate up to 72.44% for error detection, which can provide more reliable target for error correction.
Keywords :
error correction; error detection; smoothing methods; speech recognition; Chinese speech recognition error detection; Chinese syllable evaluation; N-gram based phonetic syllable evaluation; bigram model; error correction; smoothing method; trigram model; Acoustic testing; Databases; Engines; Error correction; Hidden Markov models; Large-scale systems; Libraries; Smoothing methods; Speech analysis; Speech recognition; Chinese; Error Detection; N-gram; Recognition; Speech; Syllable;
Conference_Titel :
Natural Language Processing and Knowledge Engineering, 2009. NLP-KE 2009. International Conference on
Conference_Location :
Dalian
Print_ISBN :
978-1-4244-4538-7
Electronic_ISBN :
978-1-4244-4540-0
DOI :
10.1109/NLPKE.2009.5313747
