RSS

Statistical Models for Speech and Language Processing

A basic line of research that we have been following, is the development and evaluation of novel statistical models for speech and language processing. Research topics have included:

  • Latent variable models for dimensionality reduction and missing data reconstruction, with applications to speech recognition,
  • Support Vector Machines for enhanced discrimination in speaker verification and recognition
Currently, we are very interested in maximum entropy modelling, feature selection algorithms and machine translation algorithms (for monolingual tasks).

A long term goal of this research is to develop models of speech recognition that can be related to speech production.

Key Publications
  • M. Carreira-Perpiñán and S. Renals.
    Dimensionality reduction of electropalatographic data using latent variable models.
    Speech Communication, 26:259-282, 1998.