Natural Language Processing
- I did my doctoral work in NLP; it was a NL parser called Plink
that parsed like humans.
- I got into CAs in that period, but didn't start to do
any simulations until well after my doctorate.
- I started doing CA simulations because I thought I could
use CAs to get semantics to solve the prepositional phrase
attachment problem.
- I thought people had worked CAs out, but it turns out
that they haven't.
- So, after a 7 year digression, I'm hoping to get back
to NLP with CAs.
- Palm has a FSA parser, but that's insufficient for
NLP.
- I'd like to build a parser and larger systems using CAs.
- It will necessarily be for a restricted domain initially,
as I have no idea how to learn open ended concepts.
- I might use the blocks world domain, or a video game enviroment.
- This would be a good move towards solving a lot of the standard
AI problems that beset NLP, e.g. symbol grounding and semantics.
- Scaling this up to larger domains, and even open-ended domains
could plausibly pass the Turing test.
