CS 789 THEORY SEMINAR [home]
Speaker: Frank
Oles
Affiliation: IBM Watson
Date: Monday, November 3, 2003
Title: Precedence-Inclusion
Patterns and Their Use in Information Extraction from Text
Abstract:
The points of this talk are to
introduce the theory of precedence-inclusion patterns, and to discuss how this
new theory can be concretely applied to the practical problem of finding
instances of relations in text based on small numbers of examples. From a
mathematical point of view, precedence-inclusion patterns are sets with a
strictly partially ordered set of strict partial orders, along with some
additional structure. The definition of these structures reflects how
multiple partial orders interact in a number of situations such as in text, in
images, and in video. In particular, precedence-inclusion patterns
generalize constituent structure trees familiar to computational linguists.
After outlining the basic properties of precedence-inclusion patterns, I
will present a basic result: each finite set of finite
precedence-inclusion patterns has a minimal most specific generalization that is
unique up to isomorphism. A simple demo will be incorporated into the
presentation to illustrate that this rather abstract
theory has computationally tractable ramifications.