In this latter case, it was recently shown [Steeneken & Van Leeuwen, 1995] that the hybrid system developed in WERNICKE was slightly ahead of more traditional HMM systems, when compared on both British and American English tasks, using a 20,000 word vocabulary and a trigram language model, along with the other leading European systems produced by LIMSI (France), Philips (Germany) and Cambridge University/HTK (UK). These recent results, however, are prior to the formal adjudication procedure.

These results are prior to the formal adjudication procedure.

It is indeed known that transitions are more important to speech perception than stationary parts.

Although no formal European funding has been required for this

Independently of any EC funding.

It might be that INESC does not need a SPRACHSTATION.

Due to the move of Hervé Bourlard, previously affiliated with L&H.

Due to the move of Steve Renals, previously affiliated with CUED.

See Section B for a short description of ICSI, whose contribution in WERNICKE, in terms of hardware, software and research, was really significant and substantial.

See earlier comment in WP6 justifying how the Euro-neuro-computers are more specialized.

Including 3 person-years for ICSI.

Jean-Marc Boite
Mon Dec 9 18:18:02 MET 1996