[Pipet Devel] Re: natural language processing

[Jeffrey Chang]

Hi Jeff,

Sounds like Piper's moving right along.  Congratulations!  The plans you
have in terms of NLP are really exciting, and I'm sure it's going to be an
exciting problem.  It sounds like the right thing to do to start, is to
compile a list of "templates" for actions that people can type in.  For
example:
  add <object-type> [named <name>]
  connect <object-name> to <object-name>
  ...

A good start would be to require people to type or say things that exactly
match a template.  Then, you can expand it, so that it could learn different
ways of expressing the information in the templates.

As for the Narval stuff, to be able to handle more abstract requests, the
system will have to understand the interrelationships between each of the
objects.  Do you have that functionality already in Piper?

Hmmm...  I'm starting to get carried away now.  Unfortunately, I'm not going
to have time to work more on that project for a while -- I'm trying to put
together my thesis now.  However, integrating with NLP with Piper sounds
exciting, and I'll check in from time to time to see what's going on.

Good luck!

Jeff



> From: "J.W. Bizzaro" <jeff at bioinformatics.org>
> Date: Thu, 03 May 2001 03:26:33 +0000
> To: Jeffrey Chang <jchang at SMI.Stanford.EDU>
> Cc: pipet-devel at bioinformatics.org
> Subject: natural language processing
> 
> Hi Jeff!
> 
> I read on the ISMB website that you're giving a tutorial on natural language
> processing.  It's interesting to see that, because we've been talking about
> NLP for Piper.  Basically, I've been thinking about a text-based interface
> that works like a first-person adventure game, where you give text commands to
> traverse nodes in Piper and do things with other objects at each node.  For
> example, you might say "connect terminal 1 to terminal 2".  And Piper would
> respond with text-based verifications of actions.
> 
> The neat thing would be to integrate this with voice recognition and speech
> synthesis.  Then, because Piper can use multiple interfaces (GUI and non-GUI),
> users could talk and see the results of commands graphically.
> 
> We're also collaborating with some people developing an intelligent personal
> agent called Narval.  They've been interested in connecting it to Piper to
> take advantage of the networking and the information stored in nodes.
> Interestingly, they too have thought about using NLP.  With Narval/Piper, you
> might tell it to get some information for you, in a very abstract way, and
> then have the agent find what it can, possibly even building its own networks
> to get results.  You can imagine the uses of this in bioinformatics! :-)
> 
> Anyway, I thought that I'd let you know and suggest, if you're interested in
> developing an NLP bioinformatics system, that you might want to consider
> Narval/Piper :-)
> 
> Cheers.
> Jeff
> -- 
> J.W. Bizzaro                                jeff at bioinformatics.org
> Director, Bioinformatics.org        http://bioinformatics.org/~jeff
> "As we enjoy great advantages from the inventions of others, we
> should be glad of an opportunity to serve others by any invention
> of ours; and this we should do freely and generously."
> -- Benjamin Franklin
> --
>