Coreference: An ontological semantic approach

Sara Jayne Shelton, Purdue University

Abstract

This study examines coreference resolution processes from a semantic-ontological perspective. The study outlines and describes 30 knowledge-based pragmatic mechanisms which account for many aspects of coreference assignment. Mechanisms are categorized according to object referents (15 mechanisms), event referents (5 mechanisms), and transcategorical referents (3 mechanisms). A fourth category, support mechanisms, discusses 7 processes which strengthen and/or block coreference links made by other mechanisms. In addition to examining standard coreference fare such as deixis from an ontological view, several new aspects of coreference resolution are introduced (e.g., ontological reference, collective reference, variable reference, information retrieval tags, and event coreference). The coreference resolution mechanisms described in this study are then applied to a set of four texts being used as a training corpus for a knowledge-based machine translation (KBMT) project at Computing Research Lab (CRL), New Mexico State University at Las Cruces. Mechanism interaction and performance in handling coreference tagging are discussed, and implications for KBMT system design are noted.

Degree

Ph.D.

Advisors

Raskin, Purdue University.

Subject Area

Linguistics|Computer science|Artificial intelligence

Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server
.

Share

COinS