Reference Advice for Word Sense Disambiguation

by Johnson, Ben, PhD

Abstract (Summary)
Much of the world’s knowledge is encoded in natural language. Accessing this in- formation would be invaluable for applications such as agent systems, question answering, the semantic web, expert systems, and many more. However, language is very ambiguous – each word in a natural language utterance can have a variety of meanings. Word sense disambiguation is the task of determining the dictionary (or lexical) sense of each word in a context. Knowing which sense a word represents allows us to ascertain its meaning. In this work, we introduce a new resource for WSD processing: reference advice. Reference advice uses reference resolution methods to infer potential meanings of referring expressions, such as ’he’, ’she,’ ’this,’ ’it’ and many more. We present a method for creating reference advice, and incorporating that advice into WSD processing. We also develop several system configurations which act as points of comparison for evaluation. Lastly, we discuss the impact of the success of reference advice in supporting WSD processing. We explore a shift away from pipeline text processing architectures, wherein each processing component feeds its results downstream until a final processing component presents a solution. We discuss how our system represents a step toward text processing architectures where each processing component is able to provide input to many other processing components, not just those components that are rigidly positioned downstream.
Full Text Links

Main Document: View

Bibliographical Information:

Advisor:Sergei Nirenburg

School:University of Maryland Baltimore County

School Location:USA - Maryland

Source Type:Doctoral Dissertation

Keywords:natural language understanding, reference resolution, machine learning, ontology, knowledge representation, word sense disambiguation, wordnet, ontosem, ontological semantics


Date of Publication:08/30/2013

© 2009 All Rights Reserved.