My research in speech synthesis has centred on employing synthesis as a way of testing the viability of declarative, Firthian Prosodic non-segmental phonology.

One of the central aspects of Firthian Prosodic Analysis (FPA) is its 'entirely representational' orientation. It attempts to provide a coherent, non-process, integration of phonological, morphological, syntactic and semantic information along with explicit statements of the phonetic interpretation of such information. This makes it an ideal candidate for exploration with declarative formalisms in speech synthesis.

One outcome my research work (initiated with John Coleman and Adrian Simpson, and continued with Richard Ogden and Steve Harlow) is a computational version of FPA which provides the basis of YorkTalk non-segmental speech synthesis system (funded by British Telecom Plc).

More recently I have collaborated with colleagues at the University of Cambridge and University College London on an EPSRC funded project (Prosynth) which sought to use and develop the kinds of phonological knowledge resulting from the YorkTalk research to improve naturalness in speech synthesis.

Some publications arising from this work are amongst those listed here here.

---

John Local's Home Page