Recent research suggests that syllable-length acoustic models might be more appropriate for pronunciation variation modelling than the context-dependent phones that conventional automatic speech recognisers use. In this paper, we compare the recognition performance of two types of recognisers: a conventional recogniser that only uses triphones, and an experimental recogniser that employs a mix ...

Exploiting the information about the vocal tract shape that produced the speech has been appealing to speech researchers and scientists for a long period of time. Experimental studies that included the articulatory information from physiological measurements supported the idea that this information could be useful in a number of areas of speech science and technology. However, the estimation of...

We propose an unsupervised approach for substring-based transliteration which incorporates two new sources of knowledge in the learning process: (i) context by learning substring mappings, as opposed to single character mappings, and (ii) phonetic features which capture cross-lingual character similarity via prior distributions. Our approach is a two-stage iterative, boot-strapping solution, wh...

1.2 The Input 1.2.1 Surface forms • Consists of surface phonetic forms to which the children may or may not have attached a meaning (so we can’t assume any sort of semantic bootstrapping, morphemic knowledge, etc). • We can’t assume knowledge of relationships among forms, and therefore we can’t assume knowledge of underlying forms. • This means positive data only—what forms exist—whereas phonot...

Integrating phonetic knowledge into a speech recognizer is a possible way to further improve the performance of conventional HMM-based speech recognition methods. This paper presents a cascaded architecture which consists of attribute detection and conditional random field to make use of phonetic knowledge within the phone decoding process. The attribute detection can be implemented by using an...

In this paper, we propose a model for recognition of utterances of consonant–vowel (CV) units. The acoustic–phonetic knowledge of the CV classes is incorporated in the form of constraints of a constraint satisfaction model. The model combines evidence from multiple classifiers. The significant feature of this model is that discrimination of the CV units could be enhanced by a combination of eve...

A long-standing forte of the Laboratory Phonology series has been work on phonetic implementation of phonological representations. Numerous studies in this series have elucidated the patterns of variation in the realization of phonological categories in different segmental and prosodic contexts, and such studies now provide one of the main lines of evidence about the cognitive representation of...