Transcribing broadcast news with the 1997 Abbot System

Previous DARPA CSR evaluations have focused on the transcription of broadcast news from both television and radio programmes. This is a challenging task because the data includes a variety of speaking styles and channel conditions. This paper describes the development of a connectionist-hidden Markov model (HMM) system, and the enhancements designed to improve the performance on broadcast news data. Both multilayer perceptron (MLP) and recurrent neural network acoustic models have been investigated. We asses the effect of using gender-dependent acoustic models, and the impact on the performance of varying both the number of parameters and the amount of training data used for acoustic modelling. The use of a context-dependent phone models is described, and the effect of the number of context classes is investigated. We also describe a method for incorporating syllable boundary information during search. Results are reported on the 1997 DARPA Hub-4 development test set.