Concept-Based Information Extraction

Issue title: Knowledge Extraction from Text

Guest editors: Steven L. LytinenGuest Editor

Article type: Research Article

Authors: Thompson, Paul^{; 1}

Affiliations: PRC Inc., 1500 PRC Drive, McLean, VA 22102

Note: [1] Present address: D1-66D, West Publishing Co., 610 Opperman Drive, P.O. Box 64526, St. Paul, MN 55164.

Abstract: PAKTUS is a shell used to develop natural language systems utilizing full syntactic and semantic processing to perform conceptually based information extraction. Such systems promise significant performance improvements over keyword-based extraction systems. Much work in this area has been in artificial domains, highly formatted sublanguage domains, or less formatted but still very narrowly focused domains. New approaches are required to extract information from less formatted and less narrowly focused domains. Several such systems have been developed using PAKTUS. These systems are designed to respond to the needs of message handling centers processing large volumes of highly volatile messages. PAKTUS has performed successfully in the four message understanding conferences (MUC) sponsored by the Naval Ocean Systems Center (NOSC) and DARPA during the past several years, including the most recent, MUC-4.

DOI: 10.3233/ICA-1994-1606

Journal: Integrated Computer-Aided Engineering, vol. 1, no. 6, pp. 537-546, 1994