Please use this identifier to cite or link to this item:
https://ptsldigital.ukm.my/jspui/handle/123456789/476230
Title: | Harmony search algorithm for Word Sense Disambiguation |
Authors: | Saad Adnan Abed (P69602) |
Supervisor: | Sabrina Tiun, Dr. |
Keywords: | Word sense disambiguation |
Issue Date: | 24-Mar-2015 |
Description: | Word Sense Disambiguation (WSD) is a task of determining which sense of a polysemous word (word with multiple meanings) is chosen in a particular use of the word by considering the context of the word. A sentence is considered as ambiguous if it contains polysemous word(s). Practically, any sentence that has been classified as ambiguous usually has multiple interpretations, but just one of them presents the correct interpretation. This research proposes an unsupervised approach for word sense disambiguation using harmony search (HS), one of the state-of-the-art meta-heuristic algorithms. The proposed approach has two phases. In the first phase, Stanford parser is used to parse sentences in order to get dependency relations, while WordNet is referred to determine all possible senses for each parsed word. In the second phase, HS algorithm is used to find the best interpretation for a sentence among all other interpretations. The goal of using the harmony search algorithm is to maximize the overall semantic similarity on the set of parsed words. A combination of three methods of semantic similarity and relatedness measurements i.e. adapted Lesk, Jiang and Conrath (JCN), and Gloss vector, is used to measure the similarity between the sense of the word and the sense of its dependent word according to the dependency relations. The similarity score between each pair of senses is enriched by domain disambiguation using WordNet domains. The combination of measurement methods supported by WordNet domains represents the fitness function for the harmony search algorithm. The experimental results yield 74.61% precision for nouns and 67.58% for all parts of speech in WordNet. The result is better compared to other unsupervised methods that have been proposed in the literature when tested on the same benchmark dataset namely SemCor 2.0. Thus, it can be concluded that the proposed approach is a good solution method for unsupervised word sense disambiguation.,Master/Sarjana |
Pages: | 87 |
Publisher: | UKM, Bangi |
Appears in Collections: | Faculty of Information Science and Technology / Fakulti Teknologi dan Sains Maklumat |
Files in This Item:
File | Description | Size | Format | |
---|---|---|---|---|
ukmvital_76527+SOURCE1+SOURCE1.0.PDF Restricted Access | 2.48 MB | Adobe PDF | View/Open |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.