Combining lexical chain and domain driven approaches to enhance lexical chain performance

Lee, Wei Jan (2012) Combining lexical chain and domain driven approaches to enhance lexical chain performance. Masters thesis, Universiti Malaysia Sarawak, (UNIMAS).

[img]
Preview
PDF
Combining lexical chain and domain driven approaches to enhance lexical chain performance (24pgs).pdf

Download (8MB) | Preview
[img] PDF (Please get the password from Digital Collection Development Unit, ext: 3932 / 3914)
Combining lexical chain and domain driven approaches to enhance lexical chain performance (fulltext).pdf
Restricted to Registered users only

Download (57MB)

Abstract

~ord Sense Disambiguation (WSD) is the process of identifying the meaning of the words in context of computational manner (Navigli, 2009). Every text or discourse is actually a composition of words, phrases, and sentences which tend to describe the similar topic. Therefore Morris and Hirst (1991) proposed an approach, named Lexical Chain approach to disambiguate the words by finding the relationships between words in the given text. Lexical Chain approach is originally used to exploit the lexical cohesion of a text by looking for the semantics relationships between words (relationships provided by the dictionarie~ Over the years, several researches were conducted to improve the performance of Lexical Chain by adapting different knowledge resources and different measurements. However the insufficient of process in determine the level of the semantics similarity of the relationships fonned between words based on the text restricted the perfonnance of the Lexical Chain. The ordinary Lexical Chain approach fonned the relationships between strongly related words such as car and vehicle, and at the same time, some of the abstract words are sometimes tended to be related such as ballot and resignation. The abstract relationships produce noises in the process of determining the most appropriate sense of the words. Therefore in this research is to propose a combination approach to improve the disambiguating perfonnance of Lexical Chain approach. The purpose of this research is to improve the sense identification by integrating Lexical Chain approach with Domain Driven appr?ach. This combination approach is derived from the concept of exploiting the lexical cohesion and textual coherence of any given text. The proposed combination approach will form the relationships between words by using the Lexical Chain approach and determines the semantics similarity based on textual coherence obtained by Domain Driven approach. Domain Driven approach is proposed to integrate with Lexical Chain approach because Domain Driven approach was proposed to exploit the coherence of any given text by accessing to the domain knowledge of the words in the text. The Domain Driven approach will act as the decision maker for determining the similarity of the related words that established by Lexical Chain approach. Hence the proposed framework does not only relying on the information obtained from either lexical cohesion or textual coherence, but obtains the wellness information from both approaches. The experiments had been carried out to prove the performance of the proposed combination framework. The results obtained from the experiments indicate an improvement when the Lexical Chain approach is integrated with Domain Driven approach.

Item Type: Thesis (Masters)
Additional Information: Thesis (M.Sc.) -- Universiti Malaysia Sarawak, 2012.
Uncontrolled Keywords: Lexical-functional grammar, Word Sense Disambiguation (WSD), unimas, university, universiti, Borneo, Malaysia, Sarawak, Kuching, Samarahan, ipta, education, Postgraduate, research, Universiti Malaysia Sarawak
Subjects: T Technology > T Technology (General)
Divisions: Academic Faculties, Institutes and Centres > Faculty of Computer Science and Information Technology
Depositing User: Karen Kornalius
Date Deposited: 10 Oct 2016 02:58
Last Modified: 10 Oct 2016 02:58
URI: http://ir.unimas.my/id/eprint/13772

Actions (For repository members only: login required)

View Item View Item