Abstract

Stochastic n-gram language models have been successfully applied in continuous speech recognition for several years. Such language models providemany computational advantages but also require huge text corpora for parameter estimation. Moreover, the texts must exactly reflect, in a statistical sense, the user's language. Estimating a language model on a sample that is not representative severely affects speech recognition performance. A solution tothis problem is provided by the Bayesian learning framework. Beyond the classical estimates, a Bayes derived interpolation model is proposed. Empirical comparisons have been carried out on a 10,000-word radiological reporting domain. Results are provided in terms of perplexity and recognition accuracy.

Citations

399	DW: Applied multivariate statistical analysis. 3-d edition. New Yersey - RA, Wichern
286	Interpolated estimation of Markov source parameters from sparse data - Jelinek, Mercer - 1980
56	Pattern Classi®cation and Scene Analysis - Duda, Hart - 1973
14	D.B.Rubin, “Maximum likelihood from incomplete data via the em algorithm - Dempster - 1977
3	G.Antoniol. Language modeling for e�cient beam-search - Federico, Cettolo, et al. - 1995