In this paper, we propose a method of formant prediction from pole and bandwidth data, and apply this method to automatically extract F1 and F2 values from a corpus of regional dialect variation in North America that contains 134,000 manual formant measurements. These predicted formants are shown to increase performance over the default formant values from a popular speech analysis package. Finally, we demonstrate that sociolinguistic analysis based on vowel formant data can be conducted reliably using the automatically predicted values, and we argue that sociolinguists should begin to use this methodology in order to be able to analyze larger amounts of data efficiently. Index Terms: formant prediction, corpus analysis, sociolinguistics 1.

We propose a new method for estimating Line Spectral Frequency (LSF) trajectories that uses unscented Kalman filtering (UKF). This method is based upon an iterative Expectation Maximisation (EM) approach in which LSF estimates are generated during a forward pass and then smoothed during a backward pass. The EM approach also provides re-estimated Kalman filter parameters for further forward-backward passes that improve estimation. This approach exploits the non-independence of neighbouring spectra. We estimate LSFs as they have good interpolation and quantization properties. This allows us to estimate LSF trajectories that are guaranteed to result in stable filters. We analyse noisy synthetic speech using this technique. The results compare favourably with other methods. 1.

This paper describes a method for obtaining smoothed vocal tract parameters from analysis during the closed phase of the glottis. The method is based upon Expectation Maximisation (EM) and uses Kalman-Rauch forward-backward iterations through a voiced segment, in which the speech data during excitation and open phases are excluded by treating them as ‘missing data’. This approach exploits the non-independence of neighbouring spectra and compensates for small numbers of available points, while preserving speaker-characteristic information and tracking variations in it. The vocal tract filter parameters are then used for inverse filtering the speech, thus obtaining estimates of the source excitation. The extracted excitation signal can be used to excite other sets of parameters to produce natural sounding speech. 1.

This paper reports on work developing an accent-independent lexicon for use in synthesising speech in English. Lexica which use phonemic transcriptions are only suitable for one accent, and developing a lexicon for a new accent is a long and laborious process. Potential solutions to this problem include the use of conversion rules to generate lexica of regional pronunciations from standard accents [1] and encoding of regional variation by means of keywords [2]. The latter proposal forms the basis of the current work. However, even if we use a keyword system for lexical transcription there are a number of remaining theoretical and methodological problems if we are to synthesise and recognise accents to a high degree of accuracy; these problems are discussed in the following paper.

Speech can be naturally described by phonetic features, such as a set of acoustic phonetic features or a set of articulatory features. This thesis establi shes the effectiveness of using phonetic features in phoneme recognition by comparing a recogniser based on them to a recogniser using an established parametrisation as a baseline. The usefulness of phonetic features serves as the foundation for the subsequent modelling of syllables. Syllables are subject to fewer of the context-sensitivity effects that hamper phone-based speech recognition. I investigate the different questions involved in creating syllable models. After training a feature-based syllable recogniser, I compare the feature based syllables against a baseline. To conclude, the feature based syllable models are compared against the baseline phoneme models in word recognition. With the resultant feature-syllable models performing well in word recognition, the featuresyllables show their future potential for large vocabulary automatic speech recognition.

In this paper a method of integrating a model of suprasegmental duration with a HMM-based recogniser at the post-processing level is presented. The N-Best utterance output is rescored using a suitable linear combination of acoustic log-likelihood (provided by a set of tied-state triphone HMMs) and duration log-likelihood (provided by a set of durational models). The durational model used in the post-processing imposes syllable-level elastic constraints on the durational behaviour of speech segments. Results are presented for word accuracy on the Resource Management database after rescoring, using two different syllable-like constraint units, a fixed-size N-phone window and simple (no constraint) phone duration probability scoring.

We analyze what functions as a YES response and a NO response for different yes/no questions. This problem is surprisingly complex: respondents do not always produce overt yes or no lexical items in response to a yes/no question. In addition, when respondents don't include a clear yes or no word, they may mean to communicate a clear YES or NO meaning, or something else. We find that the classification of yes/no questions described in (Carletta et al., 1995) for the Edinburgh map task corpus correlates well with whether a response will be a bare yes or no, a yes or no plus additional speech, or just speech without an overt yes or no. Correlation with responses described simply as as "direct" or "indirect" is less good. We also find that, under the three-way categorization, the strength of a question's expectation for a YES response predicts the form of the response. 1 Introduction In order to model dialogue, it is necessary to understand what kinds of utterances are likely to occur in...

This paper describes a method for using intonation to reduce word error rate in a speech recognition system designed to recognise spontaneous dialogue speech. We use a form of dialogue analysis based on the theory of conversational games. Different move types under this analysis conform to different language models. Different move types are also characterised by different intonational tunes. Our overall recognition strategy is first to predict from intonation the type of game move that a test utterance represents, and then to use a bigram language model for that type of move during recognition. 1 INTRODUCTION This paper describes a method for using intonation to reduce word error rate in a speech recognition system designed to recognise spontaneous dialogue speech. Our experiments are on the DCIEM Maptask corpus [2], a corpus of spontaneous task-oriented dialogue speech. Our dialogue analysis is based on the theory of conversational games first introduced by Power [9] and adapted for ...

Most English text-to-speech synthesisers offer one of only two accents: General American or RP. Developing a new accent is laborious, since it is not possible to choose one accent as a base form and systematically translate to others. We use the approach of Wells ([1]), categorising vowels in terms of abstract keywords that encode classes of words. Thus it is unnecessary to use a phonemic transcription in either the development or the execution of a synthesiser. The "keyvowel" system can be used throughout the synthesis system, avoiding the need to make accent-specific changes manually. The same linguistic resources can be re-used for each new accent. More fundamentally, the keyvowel system functions as a meta-accent that subsumes vowel-related information in all accents of English. 1. THE NEED FOR REGIONAL ACCENTS IN ENGLISH SPEECH SYNTHESIS A language may include several accents, differing not in their syntactic rules (as for dialects), but merely in the pronunciation rules. There a...

We present w ork intended to improve speech recognition performance for computer dialogue by taking into account the way that dialogue context and intonational tune interact to limit the possibilities for what an utterance might be.We report here on the extra constraint a c hieved in a bigram language model, expressed in terms of entropy, b y using separate submodels for dierent sorts of dialogue acts, and trying to predict which submodel to apply by analysis of the intonation of the sentence being recognised.

This paper describes a synthesis from analysis scheme for producing natural sounding intonation for speech synthesis. The paper presents a new method of describing F_0 contours in terms of three basic phonetic intonation elements. Details are given of an automatic system for labelling F_0 contours, which could be used for speech recognition purposes. Current work on extracting a phonological description from this phonetic description is discussed.