Conversational Systems and Text Extraction
- My idea when we proposed it was to do FSA based systems.
- We didn't get around to that (that's the problem with showing
up in week three).
- Chatbots based just on FSAs with some memory can be relatively
sophisticated. There's no reason these can't run on a SpiNNaker
board and remember some aspects of a conversation.
- Moreover, I thought we could do a GATE like cascade.
- One of the big things that came out of MUC was you could do
really fast processing with FSAs.
- You could cascade these togethers and do (among other things)
text extraction.
- The win is that these can all run in parallel on the neural
system.
- The bigger your cascade, the more parallelism you get.
- Maybe next year....
