Projects

For a summary of our current research, you can read our Mid-Term Report, written March 2009. The rest of the Projects page will be updated to give more detail about recent results after the spring 2010 reports are written.

S2S's research is based around 11 interrelated projects, which are grouped into 4 themes:
Theme I: Multi-linguistic and comparative research on fine phonetic detail - Projects 1-4
Theme II: Imperfect knowledge/signal - Projects 5-6
Theme III: Beyond short units of speech - Projects 7-8
Theme IV: Exemplars and abstraction - Projects 9-11

Jump to:
Project 1 | Project 2 | Project 3 | Project 4 | Project 5 | Project 6 | Project 7 | Project 8 | Project 9 | Project 10 | Project 11 |



Project 9: Hybrid episodic-abstract computational modelling: ASR

Two fundamentals of traditional HMM ASR systems, namely generalization of phoneme-like units and a top-down search procedure, are in conflict with current understanding of HSP, i.e. the preservation of FPD and the importance of bottom-up processing. Properties of recognition of novel words and L2 learning require early abstraction of (sub)-phoneme-like units; while other phenomena hint at detailed segmental matching. Project 9 seeks to further improve its recent combined episodic and abstractionist SR model. The focus for the first half of the project will be on weighting acoustic-phonetic information relative to more abstract knowledge. Later this project may seek to incorporate systematic linguistic knowledge fed to it by Project 10, using the bottom-up information in ways that humans are thought to.

View Project

Working on this project: » Prof Dirk Van Compernolle » Prof Roger Moore » Dr Sébastien Demange » Dr Kris Demuynck » Dr Dino Seppi

September 2010
S M T W T F S
      1 2 3 4
5 6 7 8 9 10 11
12 13 14 15 16 17 18
19 20 21 22 23 24 25
26 27 28 29 30    

Marie Curie Logo