WORD SENSE DISAMBIGUATION

Author:

Kumar, Saket

Graduate Program:

Computer Science

Degree:

Master of Science

Document Type:

Master Thesis

Date of Defense:

March 24, 2015

Committee Members:

Omar A El Ariss, Thesis Advisor/Co-Advisor
Jeremy Joseph Blum, Thesis Advisor/Co-Advisor
Linda Marie Null, Thesis Advisor/Co-Advisor
Sukmoon Chang, Thesis Advisor/Co-Advisor
Thang Nguyen Bui, Thesis Advisor/Co-Advisor

Keywords:

Sense
Word
Disambiguation
Bee
Lesk

Abstract:

Humans can infer meaning through the use of not only the definition of a word, but also, where one word might have various conflicting definitions, based on their experiences and the text’s context and domain. Word Sense Disambiguation (WSD) is the problem of finding the most appropriate meaning of a word in a particular context. The functional importance of WSD lies in processing the sequence of words by pinpointing their meaning without the need for human intervention. It is crucial for many applications such as translation, summarization, information retrieval, and many other natural language applications. We introduce an unsupervised knowledge-source approach for word sense disambiguation using a bee colony optimization algorithm. We also present several variations to our bee colony approach that improve the overall performance of the algorithm. Our results are compared with recent unsupervised approaches such as ant colony optimization, genetic algorithms, most frequent sense, and simulated annealing.