Context Free Parsing
- I made two CF parsers.
- The first one I made was for CABot1. It used a stack (about 5 items),
and bound with STP.
- The problem was that the timing was all wrong. It took way too long
to parse a sentence.
- I followed Rick Lewis (and Vasishth Cog Sci 2005) and switched
from a stack to active memory.
- That is, you can see how active a phrase is and use that too
simulate a stack.
- It turns out that that's a better cognitive answer than a stack.
- How do you tell how active a concept (or phrase) is? You count how
many neurons fire.
- Unfortunately, we don't have a good way to make a CA do this.
Psychological CA persistence is an open question.
- We solved it by programming it (in neurons).
