We propose a stochastic pronunciation model using an ergodic hidden Markov model (EHMM) of automatically derived acoustic sub-word units (SWU). The proposed EHMM discovers the pronunciation structure inherent in the acoustic training data of a word without any apriori phonetic transcriptions. The EHMM is an HMM of HMMs – its states are SWU HMMs and the state-transitions compose various possible...

In this paper we measure the degrees of association among aggregate pronunciational, lexical and syntactic differences in 70 Dutch dialect varieties. First, we show that pronunciation is marginally more strongly associated with syntax than it is with lexis and that syntax and lexis are only weakly associated. Then, we check for the influence of geography as an underlying factor because geograph...

A significant source of variation in spontaneous speech is due to intra-speaker pronunciation changes. Previous work in automatic speech recognition has identified several factors that affect pronunciation variability such as phonetic context and speaking rate, as well as syntactic structure. This work examines prosody as a cue to pronunciation variability, as represented by attributes derived ...

In this paper, we evaluate a set of linguistic rules for pronunciation variations in Singapore English. We collect and annotate a speech corpus for Singapore English and label it with IPA narrow transcriptions. Data driven pronunciation rules are derived using American English (Buckeye corpus) as a reference. We compare the data driven rules with linguistic rules proposed by phoneticians, and f...

This paper extends n-gram graphone model pronunciation generation to use a mixture of such models. This technique is useful when pronunciation data is for a specific variant (or set of variants) of a language, such as for a dialect, and only a small amount of pronunciation dictionary training data for that specific variant is available. The performance of the interpolated ngram graphone model i...

In this paper two di erent models of pronunciation are presented: the rst model is based on a rule set compiled by an expert, while the second is statistically based, exploiting a survey about pronunciation variants occurring in training data. Both models generate pronunciation variants from the canonic forms of words. The two models are evaluated by applying them to the task of automatic segme...

This paper proposes a new method of pronunciation variant generation for reducing word error rate in conversational speech recognition. In particular, this paper focuses on the generation of alternative pronunciations from canonical forms by using the phonological knowledge derived from the analysis of a phonetic transcription corpus. The experimental results show that the pronunciation variati...

This article reports on an action research project which investigated the value of systematically and explicitly incorporating a pronunciation sub-syllabus within the overall syllabus of a full-time post-intermediate level ESOL course. This pronunciation syllabus involved raising each individual learner’s awareness of their difficulties with pronunciation and of the main features of spoken Engl...