12:00, 12 Nov 1997, WeH 7220

		 First-Order Learning for Web Mining
			     Mark Craven


In recent years, there has been a large body of research centered
around the topic of learning first-order representations.  This work,
is appealing in that first-order representations can succinctly
represent a much larger class of concepts than can more commonly used
propositional representations.  Although learning with first-order
representations holds great promise, to date there have been only a
few problem domains in which first-order representations have
demonstrated a decided advantage over propositional representations.
The World Wide Web, however, presents interesting opportunities for
the application of first order learning methods since the Web can
naturally be viewed as a directed graph.  I will discuss the
application of first-order methods to the task of learning relations
between pages in the World Wide Web.  This task is relevant to both
information extraction and resource discovery applications.