H.; A Phone Recognizer Helps to Recognize Words Better

For most speech recognition systems dynamic features are the only way to incorporate temporal context into the out-put distributions of the HMMs. In this paper we propose an efficient method to utilize a large context in the recognition process. State scores of a phone recognizer which runs in parallel to the word recognizer are computed. Integrating these scores in the HMMs of the word recognizer makes their output densities context-dependent. The approach is evaluated on a set of spontaneous utterances which have been recorded with our spoken dialogue system. A signifi-cant reduction of the word error rate has been achieved. 1