TY - CONF
T1 - SOURCE-EXTENDED LANGUAGE MODEL FOR LARGE VOCABULARY CONTINUOUS SPEECH RECOGNITION
AU - Kobayashi, Tetsunori
AU - Wada, Yosuke
AU - Kobayashi, Norihiko
N1 - Funding Information:
In this research, we used the corpus of 1994 Mainichi newspaper articles, and its differential corpus for morpheme analysis made by RWCP (RWC-DB-TEXT-95-1) and ASJ continuous speech corpus (JNAS).
Publisher Copyright:
© 1998. 5th International Conference on Spoken Language Processing, ICSLP 1998. All rights reserved.
PY - 1998
Y1 - 1998
N2 - Information source extension is utilized to improve the language model for large vocabulary continuous speech recognition (LVCSR). McMillan's theory, source extension make the model entropy close to the real source entropy, implies that the better language model can be obtained by source extension (making new unit through word concatenations and using the new unit for the language modeling). In this paper, we examined the effectiveness of this source extension. Here, we tested two methods of source extension: frequency-based extension and entropy-based extension. We tested the effect in terms of perplexity and recognition accuracy using Mainichi newspaper articles and IN AS speech corpus. As the results, the bi-gram perplexity is improved from 98.6 to 70.8 and tri-gram perplexity is improved from Jt1.9 to 26.4- The bigram-based recognition accuracy is improved from 79.8% to 85.3%.
AB - Information source extension is utilized to improve the language model for large vocabulary continuous speech recognition (LVCSR). McMillan's theory, source extension make the model entropy close to the real source entropy, implies that the better language model can be obtained by source extension (making new unit through word concatenations and using the new unit for the language modeling). In this paper, we examined the effectiveness of this source extension. Here, we tested two methods of source extension: frequency-based extension and entropy-based extension. We tested the effect in terms of perplexity and recognition accuracy using Mainichi newspaper articles and IN AS speech corpus. As the results, the bi-gram perplexity is improved from 98.6 to 70.8 and tri-gram perplexity is improved from Jt1.9 to 26.4- The bigram-based recognition accuracy is improved from 79.8% to 85.3%.
UR - http://www.scopus.com/inward/record.url?scp=0005297498&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=0005297498&partnerID=8YFLogxK
M3 - Paper
AN - SCOPUS:0005297498
T2 - 5th International Conference on Spoken Language Processing, ICSLP 1998
Y2 - 30 November 1998 through 4 December 1998
ER -