Although I have become identified with the theory group, since that
is the primary focus of Ch. 2 of my forthcoming book, I have also been
working on natural language processing for a long time, and I realize
the need for very large lexicons and ontologies for supporting NLP.
I have been encouraged by the incipient collaboration that Ed Hovy
and I had in aligning our thematic roles. That should not be a surprise,
since both of us were tuned into the NL work that has been going on
for many years. As a result, we were able to get a combined set of
roles with a minimum amount of renaming and realignment.
Two weeks ago, I also spent a day visiting Nicola's group at CNR,
and we had some very useful discussions. But a lot more than one-day
visits are needed. Two years ago, I spent several days visiting Cyc
in Austin, and we got a better understanding of each other's goals
and approaches, but that wasn't enough time to reach a consensus.
I think that the meeting at Stanford in November will be a useful
update on what people are doing. But I doubt that it will satisfy
Nicola and Doug's desire to reach some firm agreements about theoretical
issues. Ed and I reached a rapid agreement on thematic roles because
we both started from common ground on a subset that has been actively
analyzed, studied, and implemented by a large community of linguists
and computational linguists over the past 30 years. I don't think
that some of the areas that Nicola and Doug are trying to address
have yet reached that level of stability.
John Sowa