Speech Recognizer with Dynamic Alternative Path Search and Its Performance Evaluation
For a middle-size (around 1,000 words) vocabulary speech recognition, a Finite State Automaton (FSA) language model is widely used. However, defining a FSA model with sufficient coverage and consistency requires much human effort. We already proposed a method to automatically construct a FSA language model from learning corpus by use of FSA DP matching algorithm. Experiment results show that this model attains quite high recognition correct rate for closed data, but only low rate for open data. This is mainly because a necessary path does not appear in a generated FSA. To cope with this problem, we propose a new search algorithm that allows to jump dynamically to an alternative path when speech recognition of some words seems to fail. We report experiment results and discuss the effectiveness of the algorithm.
KeywordsFSA language model Automatic construction of a language model Dynamic alternative path search Speech recognition performance
- 1.Lang, K.J., Pearlmutter, B.A., & Price, R. (1998). Results of the Abbadingo One DFA Learning Competition and a New Evidence Driven State Merging Algorithm. Proceedings of the International Colloquium Grammatical Inference, pp. 1–12.Google Scholar
- 2.Lucas, S.M., & Reynolds, T.J. (Jul 2006). Learning deterministic finite automata with a smart state labeling evolutionary algorithm., IEEE Transactions on Pattern Analysis and Machine Intelligence, 27(7).Google Scholar
- 3.Kermorvant, C., de la Hinguera, C., & Dupont, P. (2004). Learning typed automata from automatically labeled data, Journal Électronique d’Intelligence Artificielle, 6(45).Google Scholar
- 4.Hu, J., Turin, W., & Brown, M.K. (1996). Language Modeling with Stochastic Automata. Proceedings of ICSLP-1996.Google Scholar
- 6.Morimoto, T., & Takahashi, S. (2008). Automatic Construction of FSA Language Model for Speech Recognition by FSA DP-Matching. In O. Castillo et al. (Eds.), Trends in intelligent systems and computer engineering (pp. 515–524). Springer.Google Scholar
- 7.Young, S., et al. (1999). The HTK Book (for Ver. 3.0)”. http://htk.eng.cam.ac.uk/.
- 8.Kawahara, T., Lee, A., Takeda, K., Itou, K., & Shikano, K. (2004). Recent Progress of Open-Source LVCSR Engine Julius and Japanese Model Repository – Software of Continuous Speech Recognition Consortium. Proceedings of ICSLP-2004. (http://julius.sourceforge.jp/en/julius.html).
- 9.National Language Research Institute (1994). Bunrui-Goi-Hyo (Word List by Semantic Principles). Syuei-Shuppan, (in Japanese).Google Scholar