The New Jersey Institute of Technology's
Electronic Theses & Dissertations Project

Title:	Automatic categorization of abstracts through Bayesian networks
Author:	Ramirez, William
View Online:	njit-etd2000-069 (x, 88 pages ~ 5.2 MB pdf)
Department:	Department of Computer and Information Science
Degree:	Master of Science
Program:	Computer Science
Document Type:	Thesis
Advisory Committee:	Scherl, Richard B. (Committee chair) Geller, James (Committee member) Perl, Yehoshua (Committee member)
Date:	2000-01
Keywords:	Machine learning. Neural networks (Computer science). Text processing (Computer science).
Availability:	Unrestricted
Abstract:	This thesis presents a method for assigning abstracts of Artificial Intelligence papers to their area of the field. The technique is implemented by the use of a Bayesian network where relevant keywords extracted from the abstract being categorized, are entered as evidence and inferencing is made to determine potential subject areas. The structure of the Bayesian network represents the causal relationship between Artificial Intelligence keywords and subject areas. Keyword components of the network are selected from precategorized abstracts. The work reported here is part of a larger project to automatically assign papers to reviewers for Artificial Intelligence conferences. The process of assigning papers to reviewers begins by using the inference system reported here to derive Artificial Intelligence subject areas for such papers. Based on those subjects, another module can select reviewers according to their specialization and limited by conflicts of interest.