Large Vocabulary Recognition of Wall Street Journal Sentences at Dragon Systems

Large Vocabulary Recognition of Wall Street Journal Sentences at Dragon Systems James Baker, Janet Baker, Paul Bamberg, Kathleen Bishop, Larry Gillick, Vera Helman, Zezhen Huang, Yoshiko Ito, Stephen Lowe, Barbara Peskin, Robert Roth, Francesco Scattone Dragon Systems, Inc. Newton, Massachusetts 02160 ABSTRACT In this paper we present some of the algorithm improvements that have been made to Dragon's continuous speech recog- nition and training prograxns, improvements that have more than halved our error rate on the Resource Management task since the last SLS meeting in February 1991. We also report the "dry run" results that we have obtMned on the 5000-word speaker-dependent Wall Street Journal recognition task, and outline our overall research strategy and plans for the future. In our system, a set of output distributions, known as the set of PELs (phonetic elements), is associated with each phoneme. The HMM for a PIC (phoneme-in-context) is rep- resented as a linear sequence of states, each having an out- put distribution chosen from the set of PELs for the given phoneme, and a (double exponential) duration distribution. In this paper we report on two methods of acoustic modeling and tr~ning. The first method involves generating a set of (unimodal) PELs for a given speaker by clustering the hypo- thetical frames found in the spectral models for that speaker, and then constructing speaker-dependent PEL sequences to represent each PIC. The "spectral model" for a PIC is sim- ply the expected value of the sequence of frames that would be generated by the PIC. The second method represents the probability distribution for each parameter in a PEL as a mixture of a fixed set of unimodal components, the mixing weights being estimated using the EM algorithm. In both models we assume that the parameters axe statistically inde- pendent. We report results obtained using each of these two meth- ods (RePELing/Respelling and univariate "tied mixtures") on the 5000-word closed-vocabulary verbalized punctuation version of the Wall Street Journal task. 1. INTRODUCTION This paper presents "dry run" results of work done at Dragon Systems on the Wall Street Journal (WSJ) benchmark task. After we give a brief description of our continuous speech recognition system, we describe the two different kinds of acoustic models that were used and explain how they were trained. Then we present *This work was sponsored by the Defense Advanced Research Projects Agency and was monitored by the Space and Naval War- fare Systems Command under contract N00039-86-C-0307. and discuss the results obtained so far and review our plans for further research. In our system a set of output distributions, known as the set of PELs (phonetic elements), is associated with each phoneme. The HMM for a PIC (phoneme-in-context) is represented as a linear sequence of states, each hav- ing an output distribution chosen from the set of PELs for the given phoneme, and a (double exponential) dura- tion distribution. The model for a particular hypothesis is constructed by concatenating the necessary sequence of PICs, based on the specified pronunciation (sequence of phonemes) …

Content truncated to 500 words