1. UKM Learning and Research Repository
2. Master Thesis / Tesis Master
3. Faculty of Information Science and Technology / Fakulti Teknologi dan Sains Maklumat

Title:	Hybridizing particle swarm optimization with simulated annealing for word sense disambiguation
Authors:	Wafaa Riyadh Musa (P75877)
Supervisor:	Sabrina Tiun, Dr.
Keywords:	Universiti Kebangsaan Malaysia -- Dissertations Dissertations, Academic -- Malaysia Words sense disambiguation Optimization methods Algorithm
Issue Date:	19-Jun-2018
Description:	Words sense disambiguation (WSD), is a task of assigning the most appropriate sense to words is used in a sentence, when the word has multiple meanings. WSD relies on the context of the target word to identify suitable sense. Selecting semantic evaluator by using an optimization strategy, is the intermediate objective to identify the set of suitable senses. Optimization methods are either based on population of solutions or single solution. However, achieving an effective balance between exploration and exploitation is a challenging task in the optimization process. Hence, this study aims to improve partial disambiguation of a sentence and find global meaning for a given text. Therefore, hybridizes a population based on an algorithm named Particle Swarm Optimization with a local search algorithm called simulated annealing algorithm (SA). PSO provides a global search of the problem space that can find various solutions of different qualities. While, the local search algorithm works on intensifying the search locally, where, promising solution is processed in this algorithm to be improved by searching its neighborhood. The hybridized method evaluates the solutions based on the semantic relation among the words. In this study, the semantic relatedness and similarity methods, which are Extended Lesk’s algorithm(e-Lesk) and Jiang-Conrath algorithm (JCN), are combined. The designed model in this research was experimented based on semantic concordance corpus (SemCor). Specifically, 19 files from this dataset, which have been used in the related works, as a benchmark dataset. Some of the related works presented their results based on only noun part-of-speech, and thus, this study did a comparison on only noun part-of-speech. While, the other comparison was based on all part-of-speeches. The proposed method outperformed other methods regarding the noun part-of-speech with f-measure of 73.36% (with increasing 0.24%). On all part-of-speech, the proposed method outperformed only at the precision metric with the highest result of 67.44% (0.41% improvement). Hence, it can be concluded that the proposed method be able to provides a good WSD solution for noun part-of-speech especially, as well as for other part-of-speech when not all words are required to be disambiguated.,Certification of Master's / Doctoral Thesis" is not available
Pages:	73
Publisher:	UKM, Bangi
Appears in Collections:	Faculty of Information Science and Technology / Fakulti Teknologi dan Sains Maklumat

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.