Dr. Eugene Charniak,
Computer Science and Cognitive Science
Brown University
Generative Models of Discourse
ABSTRACT:
Traditionally the "meaning" aspect of language is divided into "semantics", how the meaning of a sentences is built up out of the meaning of the words, and "discourse", how the meaning of an entire text (we will be discussing news articles) is built up out of the meaning of the sentences.
As is almost always the case in the study of human abilities our understanding of what is going on in discourse is tremendously limited. We thus look at the simpler problems in the hope that we can gain some purchase on the more complicated ones. In this spirit, consider the following comparatively clear cut problem. Take a news article and permute the order of the sentences. I then give you (or my program) the original and the randomly permuted version and ask you (or it) to tell which is which. Obviously, if you can, it must be on the basis of discourse knowledge, since the individual sentences remain unperturbed.
In this talk we present four generative models, each taking more discourse information into account, and each performing better than the last at our problem. Finally we discuss what use any of this might be.
BIO: Eugene Charniak is University Professor of Computer Science and Cognitive Science at Brown University and past chair of the Department of Computer Science. He received his A.B. degree in Physics from University of Chicago, and a Ph.D. from M.I.T. in Computer Science. He has published four books the most recent being Statistical Language Learning. He is a Fellow of the American Association of Artificial Intelligence and was previously a Councilor of the organization. His research has always been in the area of language understanding or technologies which relate to it. Over the last 15 years he has been interested in statistical techniques for many areas of language processing including parsing and, most recently, discourse.
Back to Distinguished Lectures | Back to News & Events