Lexical Access with a Statistically-Derived Phonetic Network

A probabilistic approach to lexieal access from a recognized phone sequence is presented. Lexical access is seen as finding the word sequence that maximizes the lexical likelihood of a sequence of phones and durations as recognized by a phone recognizer. This is theoretically correct for minimum error rate recognition within the model presented and is intuitively pleasing since it means that the "confusion matrix" of the phone recognizer will be learned and its regularities exploited. The lexical likelihoods are estimated from training data provided by the phone recognizer using statistical decision trees. Classification trees are used to estimate the phone realiziation distributions and regression trees are used to estimate the phone duration distributions, We find they can capture effectively allophonic variation, alternative pronunciation, word co-articulation and segmental durations. We describe a simpified, but efficient implementation of these models to lexical access in the DARPA resource management recognitiion task. 1. I N T R O D U C T I O N We describe a new approach to lexical access in a phonebased speech recognition system. By "lexical access" we mean taking a sequence (or, more generally, a lattice) of phones and durations that is output by a phone recognizer and mapping it onto a word sequence (or, more generally, a lattice). In conventional word-based speech recognizers, segmental durations, word co-articulation and alternative pronunciations are usually poorly modelled if at all since the architecture is not convenient or efficient for exploiting these constraints. Phone-based recognition offers an attractive alternative from this point of view. Our approach will be to create a probabilistic model that provides the likelihood that a particular word sequence gives rise to a particular phone sequence. This model will take into account allophonic variation, alternative pronunciation, word co-articulation and segmental durations. We then combine these lexical likelihoods with the acoustic likelihoods generated by the phone recognizer and priors from our language model to get an overall recognition model whose error rate we seek to minimize. We have taken this stochastic approach for two reasons. First, it provides a principled way to combine seemingly disparate information: (a) acoustic likelihoods, (b) segmental durations, (c) alternative pronunciations, and (d) the language model. Second, the availability of large speech corpora now allow the statistical estimation of these probabilities. 2. P R O B A B I L I S T I C M O D E L We form the probabilistic model as follows. Let w be a sequence of words, let y be a sequence of phones, let d be a sequence of durations, and let s be a (fixed) speech signal. Then P(wis ) ~x Z P ( s l y , d ) P ( y , d [ w ) P(w). (2.1) y,d The lefthand side of this relation is the probability that a given speech signal corresponds to a particular word sequence. The word sequence that maximizes this term gives the minimum sentence error rate. The first factor on the righthand side gives the acoustic likelihoods provided by the phone recognizer. The second factor gives the lexical likelihoods to be provided by the lexlcal access stage describe here. The third factor represents whatever language model we use. In this paper, we have used the output of the current Bell Labs phone recognizer as input to the lexical access component [1]. At present, this recognizer outputs a single sequence of phones and durations per utterance, which represents its best estimate of the true sequence. As such, y and d are fixed in Eq. 2.1 for a given speech signal. A more general approach, which would consider alternative sequences phone lattices is currently under investigation, but not reported here. Also in this paper, in which we present results on the DARPA resource management task, we consider only the the simple wordpair language model. Thus, for a given utterance, the best scoring word sequence, w, will be the one that maximizes the lexical likelihood, P(y , d[w) for a given phone recognizer output y and d, and which is a legal sequence in the word-pair grammar. In this model, finding the word sequence that maximizes this likelihood is the goal of lexical access and estimating this likelihood is the goal of this paper. A crucial factor for this estimation is that y and d are not the true sequence of phones and durations, but the output of a phone recognizer. As such, we must train our estimator on the output of the phone recognizer. This is theoretically correct for minimum error rate recognition in tnis model and is intuitively pleasing since it means that the model will learn the "confusion