I've looked through your emails and Alex's web pages. It looks
good to me. Here are a few points for what they're worth:
1. When Andrew Slater and Nick Youd worked on a parser for Infovox, they
chose what they described as a relatively flat system, and used Bachenko
and Fitzpatrick (1990) Computational Linguistics 16(3) as a springboard.
Are you also going for a flat structure (not as rich as for recognition,
e.g.), or are there reasons not to?
2. Mark's document -- very positive reaction: just 2 points
(a) paragraph 7: An algorithm searching for spectral similarity could be
useful, but I think we need a formalism to address asynchronous points
too. If there's a formal phone onset time, then subphone aysnchronies can
be formally defined as 0 plus or minus n sampling units.
(b) Paragraph 8: We need a 4-5 ms sampling frame to do consonants.
Sarah
______________________________________________________________________
Dr. Sarah Hawkins Email: sh110@cam.ac.uk
Dept. of Linguistics Phone: +44 1223 33 50 52
University of Cambridge Fax: +44 1223 33 50 53
Sidgwick Avenue or +44 1223 33 50 62
Cambridge CB3 9DA
United Kingdom