A Dynamic In-Search Discriminative Training Approach for Large Vocabulary Speech Recognition

In this paper, we propose a dynamical in-search discriminative training approach for a large-scale HMM model in large vocabulary speech recognition. A previously proposed data selection method is used to choose competing hypotheses dynamically during Viterbi beam search procedure. Particularly, all active work-ending paths are examined during search with reference transcription to identify competing tokens for different HMM's. Then HMMs are re-estimated based on a GPD-like discriminative training approach to minimize total number of possible error tokens among all collected competing tokens. In this way, recognition errors, e.g., word error rate, in training data can be reduced indirectly. The proposed approach is flexible enough to run in a batch or incremental mode. Also, the method can efficiently be implemented to process large amount of training data and update a large- scale state-tied HMM set for large vocabulary tasks. Some preliminary results on DARPA communicator task show the new discriminative training method can improve recogntion performance over our best ML-trained system.