Online LDA-Based Language Model Adaptation
In this paper, we present our improvements in online topic-based language model adaptation. Our aim is to enhance the automatic speech recognition of a multi-topic speech which is to be recognized in the real-time (online). Latent Dirichlet Allocation (LDA) is an unsupervised topic model designed to uncover hidden semantic relationships between words and documents in a text corpus and thus reveal latent topics automatically. We use LDA to cluster the text corpus and to predict topics online from partial hypotheses during the real-time speech recognition. Based on detected topic changes in the speech, we adapt the language model on-the-fly. We are demonstrating the improvement of our system on the task of online subtitling of TV news, where we achieved \(18\%\) relative reduction of perplexity and \(3.52\%\) relative reduction of WER over non-adapted system.
KeywordsTopic modeling Language model adaptation
This paper was supported by the project no. P103/12/G084 of the Grant Agency of the Czech Republic and by the grant of the University of West Bohemia, project no. SGS-2016-039.
- 2.Chen, L., Lamel, L., Gauvain, J.L., Adda, G.: Dynamic language modeling for broadcast news. In: Eighth International Conference on Spoken Language Processing (2004)Google Scholar
- 4.Tam, Y.C., Schultz, T.: Dynamic language model adaptation using variational bayes inference. In: Ninth European Conference on Speech Communication and Technology (2005)Google Scholar
- 5.Hsu, B.J.P., Glass, J.: Style & topic language model adaptation using HMM-LDA. In: Proceedings of the 2006 Conference on Empirical Methods in Natural Language Processing, pp. 373–381. Association for Computational Linguistics (2006)Google Scholar
- 6.Heidel, A., Chang, H., Lee, L.: Language model adaptation using latent Dirichlet allocation and an efficient topic inference algorithm. In: Eighth Annual Conference of the International Speech Communication Association (2007)Google Scholar
- 7.Liu, Y., Liu, F.: Unsupervised language model adaptation via topic modeling based on named entity hypotheses. In: IEEE International Conference on Acoustics, Speech and Signal Processing 2008, ICASSP 2008, pp. 4921–4924. IEEE (2008)Google Scholar
- 8.Haidar, M.A., O’Shaughnessy, D.: Unsupervised language model adaptation using latent Dirichlet allocation and dynamic marginals. In: 2011 19th European Signal Processing Conference, pp. 1480–1484. IEEE (2011)Google Scholar
- 9.Jeon, H.B., Lee, S.Y.: Language model adaptation based on topic probability of latent dirichlet allocation. ETRI J. 38(3), 487–493 (2016)Google Scholar
- 10.Hoffman, M., Bach, F.R., Blei, D.M.: Online learning for latent Dirichlet allocation. In: Advances in Neural Information Processing Systems, pp. 856–864 (2010)Google Scholar
- 11.Pražák, A., Loose, Z., Trmal, J., Psutka, J.V., Psutka, J.: Novel approach to live captioning through re-speaking: tailoring speech recognition to re-speaker’s needs. In: INTERSPEECH (2012)Google Scholar