Third International Conference on Spoken Language Processing (ICSLP 94)

Yokohama, Japan
September 18-22, 1994

One-Pass Continuous Speech Recognition Directed by Generalized LR Parsing

Kenji Kita (1), Yoneo Yano (1), Tsuyoshi Morimoto (2)

(1) Faculty of Engineering, Tokushima University, Tokushima, Japan
(2) ATR Interpreting Telecommunications Research Lab., Kyoto, Japan

In this paper, we will formulate a novel continuous speech recognition algorithm that integrates (1) a generalized LR parser for handling context-free grammar (CFG) constraints, and (2) the one-pass search algorithm for achieving an efficient search. Our algorithm enables us to efficiently find an optimal hypothesis for a given speech signal according to a specified CFG in a frame-synchronous process. Although some CFG-based speech recognition algorithms have been proposed, our algorithm is very efficient because of the introduction of three novel techniques: (1) LR path merging, (2) shared tree-structured stack, and (3) LR-parser-based dynamic network generation. We conducted an experimental comparison between our algorithm and the following two methods: (1) the one-pass search algorithm using the finite-state approximation for a CFG, and (2) the HMM-LR algorithm. Our experiments showed that our algorithm outperformed the other two algorithms.

Full Paper

Bibliographic reference.  Kita, Kenji / Yano, Yoneo / Morimoto, Tsuyoshi (1994): "One-pass continuous speech recognition directed by generalized LR parsing", In ICSLP-1994, 13-16.