Parsing

We've done some natural language parsing work with simulated spiking neurons over the years.
We wrote this in PyNN and in runs on the Nest neural simulator and SpiNNaker neuromorphic hardware.
It's part of the Human Brain Project, and PyNN is the middleware for supporting multiple hardware platforms.
We did a simple regular grammar parser this year at the Telluride Neuromorphic Workshop.
We use Finite State Automatas.
States were just well connected sets of neurons.
Once activated, they persisted indefinitely, until explicitly turned off (e.g. by the next state).
The bottom 50 neurons are parse states, and those above are semantic states. The first 200 ms. represents "John is in the kitchen." , the second "Chris is in the library." and the third just the initial state.