Re: Hierarchical Annotation Proposals

Sarah Hawkins (sh110@cam.ac.uk)
Tue, 07 Oct 1997 11:22:10 +0100

Not sending this to prosynth for threading, since Jill's use of that route
failed (eye of needle blocked?)

Alex: If it's any help, as far as I remember from my distinctly hands-off
supervision of Andrew and Nick's work, they ended up with something quite
a bit richer than MARSEC, and I think they did their own tagger, so
presumably they thought the LOB one was deficient in some way. Jill has
just asked me about the status of their final report, which I have a copy
of. Our contract with IVX was that the knoweldge was ours, but the
detailed solutions were not. I;ve suggested that, since Jill was actively
involved in some of Nick and ANdrew's work, she work out how to legally
use her knowledge. She'll be getting back to me/us on that. I'm happy to
release the report if it's legal to do so, but not otherwise of course. If
Jill decides we can;t use the report, we could legally pick Nick and
Andrew's brains.

Jill: If you end up unclear whether to read the report or not, I think
Bengt would be quite happy to get a request from you.

hope this is helpful

Sarah

On Mon, 6 Oct 1997, Alex Chengyu Fang wrote:

> At 11:51 AM 6/10/97, Sarah Hawkins wrote:
>
> >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?
>
> What I have on the web page is a description of the formal grammar my parser
> uses. It is not necessarily what is going to be put in the database as the
> syntactic description for the project. I don't see why we shouldn't use a
> flat structure if people agree that it is sufficient for the project.
>
> I've been wondering if we should try to be compatible with other schemes
> such as MARSEC, which is annotated with a very flat structure together with
> the LOB tagset. However, my experience with SpeechMaker is that a rich
> scheme, such as the one described in my web page, is a useful resource to
> draw on if ever the synthesiser needs good clausal and inter-phrasal
> descriptions.
>
> Alex
>
>