Natural Language Parser

For CABot1 I developed a stack based parser.
Binding was done via STP.
Rules were activated by the contents of the stack.
I then noticed some of Rick Lewis' work on activation based parsing in ACT.
The idea is that the stack is formed by the activity level of the item, and the activity level goes down after it is used.
As the timing was all wrong for the stack based mechanism, I implemented the activation level.
Activation level for a CA (word or phrase) was maintained by a set of of neurons that decayed naturally once activated.
Putting these things together, the parser produced correct semantic results, parsed in correct time, and resolved prepostional phrase attachment ambiguity correctly.
There are better symbolic cognitive models of parsing, but nothing neural really approaches this.