Tuesday, April 25, 2006

[Reading] Scripts, Plans, Goals and Understanding, Chapter 1,2

So from the first two chapters of Schank's "Scripts, Plans, Goals, and Understanding", I found a couple things, and I make a summary right here.

Basically, Schank's Conceptual Dependency (CD) provides eleven different kinds of ACT, namely ATRANS, PTRANS, PROPEL, MOVE, GRASP, INGEST, EXPEL, MTRANS, MBUILD, SPEAK, and ATTEND. Using these acts, one can analyze a story and make a CD representation, in which states, mental states, and acts are linked through causal links such as "result in", enable", "initiate (mental states)", "is the reason for", "disable", "result in a state that enables", and "initiate a thought that is the reason for".

I haven't read the chapter about scripts in this book, but I think we may use existing knowledge to help understanding stories. In order to do this, the existing knowledge has to be in the CD representation form, such that when we translate the natural language stories into CD representation, existing knowledge will be able to be filled in the gaps during the reasoning process. So the hard question is, how are we gonna build a knowledge base that uses CD as a representation, using OMCS as a tool/resource? Or, is it possible for us to use CD knowledge built with OMCS as a tool to translate all the stories that can be found online into CD format, and make a even larger knowledege base ("script" base, if using Schank's terminology)?

The other thing I've been thinking about is, if we map Schank's CD stuff into Marvin's emotion machine model, is it gonna be the case that different ACTs shall go to different levels, and that we're gonna find the missing ACTs when we consider how the mind works in terms of reasoning and memorizing stories?

0 Comments:

Post a Comment

<< Home