Style-Specific Language Model Adaptation using TF*IDF Similarity for Korean Conversational Speech Recognition

  • Published : 2004.06.01

Abstract

In this paper, we propose a style-specific language model adaptation scheme using n-gram based tf*idf similarity for Korean spontaneous speech recognition. Korean spontaneous speech shows especially different style-specific characteristics such as filled pauses, word omission, and contraction, which are related to function words and depend on preceding or following words. To reflect these style-specific characteristics and overcome insufficient data for training language model, we estimate in-domain dependent n-gram model by relevance weighting of out-of-domain text data according to their n-. gram based tf*idf similarity, in which in-domain language model include disfluency model. Recognition results show that n-gram based tf*idf similarity weighting effectively reflects style difference.

Keywords

References

  1. R. Iyer and M. Ostendorf, 'Relevance weighting for combining multidomain data for Ngram language modeling,' Computer Speech and Language, Vol. 13, pp. 267-282, 1999 https://doi.org/10.1006/csla.1999.0124
  2. Y.-H. Park and M. Chung, 'Analysis of Korean spontaneous speech characteristics for spoken dialogue recognition,' Journal of The Acoustical Society of Korea, vol 21, no 3, pp.330-338, 2002
  3. M. Mahajan, D. Beeferman and X. D. Huang, 'Improved topicdependent language modeling using information retrieval techniques,'Proc. ICASSP, vol 1, pp. 541-544, 1999
  4. A. Stolcke and E. Shriberg, 'Statistical language modeling for speech disfluencies,' Proc. ICASSP, vol, 1, pp,405-408, 1996
  5. M. Siu and M. Ostendorf, 'Modeling disfluencies in conversational speech,' Proc. ICSLP, vol 1, pp.621-625, 1996
  6. D.-H. Ahn and M. Chung, 'Compact subnetwork based large vocabulary continuous speech recognition,' Proc. ICSLP, vol, 1, pp.25-728, 2002