Fuzzy semantic classifier for determining strength levels of customer product reviews

Opinion Mining (OM) is one of the new paradigms of information retrieval and computational linguistics. This paradigm is not only concerned with document topic but also the opinion which is expressed. The most challenging area in OM is finding the orientation of customer feeling in reviews such as...

Full description

Saved in:

Bibliographic Details
Main Author:	Nadali, Samaneh
Format:	Thesis
Language:	English
Published:	2012
Subjects:	Fuzzy systems Semantic computing
Online Access:	http://psasir.upm.edu.my/id/eprint/32229/1/FSKTM%202012%2011R.pdf
Tags:	Add Tag No Tags, Be the first to tag this record!

id	my-upm-ir.32229
record_format	uketd_dc
spelling	my-upm-ir.322292015-01-19T07:01:37Z Fuzzy semantic classifier for determining strength levels of customer product reviews 2012-08 Nadali, Samaneh Opinion Mining (OM) is one of the new paradigms of information retrieval and computational linguistics. This paradigm is not only concerned with document topic but also the opinion which is expressed. The most challenging area in OM is finding the orientation of customer feeling in reviews such as blogs, product reviews and so on. Opinion about products is nowadays available from blogs and review sites. So, extracting opinion from these reviews help the user as well merchants to track the most suitable choice of product. There are various tasks in OM. Classification of customer reviews into positive, negative and neutral classes (also known as semantic classification) is one of the tasks that help product manufacturers or businesses to easily identify orientation of their product services. Previous studies focused on the automatic identification of opinion i.e. classifying reviews into positive, negative and neutral only. However, for some applications like flame detection or information analysis, recognizing opinion only might not be sufficient. Thus, identifying strength of opinion is considered as one of the propounded problems from the early days. In this thesis, we extended the holistic lexicon-based approach to opinion mining presented in (Ding et al., 2008), in which the researcher did not focus on finding the strength levels of opinion of each product reviews. To address the mentioned problem, a Fuzzy Semantic Classifier (FSC) is proposed to identify semantic orientation of customer product reviews at a granularity levels such as very strong, strong, moderate, weak, and very weak for each positive and negative class by combining opinion words (i.e. adverb, adjective, verb, and noun). We used fuzzy logic as it is not only using non-numerical values but also it introduces the notion of linguistic variables to overcome the uncertainty of natural language. The proposed classifier (FSC) has been tested on eight benchmark datasets introduced by (Ding et al., 2008). The results of the study showed that a Fuzzy Semantic Classifier (FSC) gave various strength of levels classification in customer product reviews which leads to multi understandability of customer opinions. The percentage of similarity between FSC and human classifications is 74%. This means that the FSC is able to classify various strength levels to very strong, strong, moderate, weak and very weak for each positive and negative class similar to human. Fuzzy systems Semantic computing 2012-08 Thesis http://psasir.upm.edu.my/id/eprint/32229/ http://psasir.upm.edu.my/id/eprint/32229/1/FSKTM%202012%2011R.pdf application/pdf en public masters Universiti Putra Malaysia Fuzzy systems Semantic computing Faculty of Computer Science and Information Technology
institution	Universiti Putra Malaysia
collection	PSAS Institutional Repository
language	English
topic	Fuzzy systems Semantic computing
spellingShingle	Fuzzy systems Semantic computing Nadali, Samaneh Fuzzy semantic classifier for determining strength levels of customer product reviews
description	Opinion Mining (OM) is one of the new paradigms of information retrieval and computational linguistics. This paradigm is not only concerned with document topic but also the opinion which is expressed. The most challenging area in OM is finding the orientation of customer feeling in reviews such as blogs, product reviews and so on. Opinion about products is nowadays available from blogs and review sites. So, extracting opinion from these reviews help the user as well merchants to track the most suitable choice of product. There are various tasks in OM. Classification of customer reviews into positive, negative and neutral classes (also known as semantic classification) is one of the tasks that help product manufacturers or businesses to easily identify orientation of their product services. Previous studies focused on the automatic identification of opinion i.e. classifying reviews into positive, negative and neutral only. However, for some applications like flame detection or information analysis, recognizing opinion only might not be sufficient. Thus, identifying strength of opinion is considered as one of the propounded problems from the early days. In this thesis, we extended the holistic lexicon-based approach to opinion mining presented in (Ding et al., 2008), in which the researcher did not focus on finding the strength levels of opinion of each product reviews. To address the mentioned problem, a Fuzzy Semantic Classifier (FSC) is proposed to identify semantic orientation of customer product reviews at a granularity levels such as very strong, strong, moderate, weak, and very weak for each positive and negative class by combining opinion words (i.e. adverb, adjective, verb, and noun). We used fuzzy logic as it is not only using non-numerical values but also it introduces the notion of linguistic variables to overcome the uncertainty of natural language. The proposed classifier (FSC) has been tested on eight benchmark datasets introduced by (Ding et al., 2008). The results of the study showed that a Fuzzy Semantic Classifier (FSC) gave various strength of levels classification in customer product reviews which leads to multi understandability of customer opinions. The percentage of similarity between FSC and human classifications is 74%. This means that the FSC is able to classify various strength levels to very strong, strong, moderate, weak and very weak for each positive and negative class similar to human.
format	Thesis
qualification_level	Master's degree
author	Nadali, Samaneh
author_facet	Nadali, Samaneh
author_sort	Nadali, Samaneh
title	Fuzzy semantic classifier for determining strength levels of customer product reviews
title_short	Fuzzy semantic classifier for determining strength levels of customer product reviews
title_full	Fuzzy semantic classifier for determining strength levels of customer product reviews
title_fullStr	Fuzzy semantic classifier for determining strength levels of customer product reviews
title_full_unstemmed	Fuzzy semantic classifier for determining strength levels of customer product reviews
title_sort	fuzzy semantic classifier for determining strength levels of customer product reviews
granting_institution	Universiti Putra Malaysia
granting_department	Faculty of Computer Science and Information Technology
publishDate	2012
url	http://psasir.upm.edu.my/id/eprint/32229/1/FSKTM%202012%2011R.pdf
_version_	1747811651590029312

Fuzzy semantic classifier for determining strength levels of customer product reviews

Similar Items