Natural Language Parser

We are hoping to develop a neurally and psycholinguistically valid natural language parser.
It does need to work on our specific sentences (17 for CABot2).
We have developed a system which is a relatively good start.
Like the CABot1 parser it produces semantic frames.
On the sentences we have tried it always succeeds.
It uses 5 lexical categories (N, V, det, period, adj).
There are slots for about 500 words (currently using ~50).
It doesn't use a stack but is Memory Based (Lewis) with active items decaying over time.
This is done by a kludge on the instance CAs.
See the improved cell assembly .
Timing is about right psycholingustically.
PP attachment ambiguity is resolved by hierarchy.
This takes advantage of hierarchies to resolve sparseness of data problems.
Incidently, my original idea for resolving the PP attachment problem got implemented by a BSc student of mine who is now a PhD student. It uses a lattice combining hierarchies from the V, N and N. It gets the best result we know of. It's similar to the neural solution
This is not a complete parser but is functional.
We hope to hook it up to grounded semantics.
We'd like to use it learn rules.
We'd like to hook it into a fuller natural language processing system that learns phonemology, morphology, and speech.