30 November: fyi -- phd funding, UK

Index of November 2007 | Index of year: 2007 | Full index





CHARACTERISATION OF REGIONAL ACCENTS OF BRITISH ENGLISH

FOR SPEECH TECHNOLOGY



Professor Martin Russell

Department of Electronic, Electrical and Computer Engineering

University of Birmingham

Birmingham, B15 2TT

UK.



Funding is available for a PhD project in the area of "Characterisation
of Regional

Accents of British English for Speech Technology" in the Department of
Electronic,

Electrical and Computer Engineering, University of Birmingham. The
funding is

sufficient to cover EU fees plus maintenance.



If you are interested in this opportunity please contact Professor
Martin Russell

(m.j.russell@bham.ac.uk)



Summary of project



The purpose of this project is to develop new mathematical models which
characterise the

acoustic properties of regional accents of British English and the
relationships between them,

and to investigate the application of these models to accent-robust
speech recognition, accent

identification, and non-native accent detection. The project will use
the existing "Accents of

the British Isles" (ABI-1 and ABI-2) corpora of recordings of read and
spontaneous

conversational speech. It will move on to consider accented telephone
conversational speech

as suitable corpora become available. The project should address the
following issues:



(1) Measurement of human accent identification performance, and the
effect of accent on

human speech recognition of read and spontaneous speech. Subjective
assessment of the

'strengths' of the accents of speakers in the ABI corpora.



(2) Application of conventional language identification techniques to
the problem of accent

identification.



(3) Automatic speech recognition (ASR) experiments to measure the effect
of accent on ASR

performance, and to measure the utility of conventional adaptation
schemes for accent

adaptation.



(4) Development of self-organising topological maps of accent data. The
objective is to

understand relationships between acoustic representations of accent, and
to see how this

correlates with geographical or historical relationships.



(5) Accent ID using topological maps.



(6) Accent adaptation of acoustic models for ASR.



(7) Accent adaptation of pronunciation dictionaries for ASR



(8) Detection of non-native-English accents in the context of general
British English. The

goal is to study how non-native speakers are mapped onto the topological
maps described

above, and in particular to determine whether it is possible to detect
non-native speakers

reliably.



(9) Characterisation of 'compound' accents: Most talkers' speech is not
'single-accent'.

Instead their voice is influenced by a number of different regional
accents. The purpose of

this task is to determine whether it is possible to detect these
compound accents and

decompose them into their component streams







Professor Martin Russell

Department of Electronic, Electrical & Computer Engineering

University of Birmingham

Birmingham

B15 2TT

UK



Tel: +44 (0) 121 414 3093

Email: m.j.russell@bham.ac.uk

Index of November 2007 | Index of year: 2007 | Full index