Sarcasm detection and classification to support sentiment analysis : A study in malay social media

The classification of users' sentiment from social media data can be used to determine public opinion on certain issues. The presence of sarcasm in text may hamper the performance of sentiment analysis. This thesis presents research work conducted on sarcasm detection and classification to supp...

Full description

Saved in:

Bibliographic Details
Main Author:	Mohd Suhairi Md Suhaimin
Format:	Thesis
Language:	English English
Published:	2017
Subjects:	HM711-806 Groups and organizations
Online Access:	https://eprints.ums.edu.my/id/eprint/37665/1/24%20PAGES.pdf https://eprints.ums.edu.my/id/eprint/37665/2/FULLTEXT.pdf
Tags:	Add Tag No Tags, Be the first to tag this record!

id	my-ums-ep.37665
record_format	uketd_dc
spelling	my-ums-ep.376652023-11-29T02:00:37Z Sarcasm detection and classification to support sentiment analysis : A study in malay social media 2017 Mohd Suhairi Md Suhaimin HM711-806 Groups and organizations The classification of users' sentiment from social media data can be used to determine public opinion on certain issues. The presence of sarcasm in text may hamper the performance of sentiment analysis. This thesis presents research work conducted on sarcasm detection and classification to support sentiment analysis. A Malay social media dataset, specifically focused on economic and political domain, was acquired from public comments posted on Facebook. The proposed work consists of two phases: (i) sarcasm detection and (ii) sentiment analysis with sarcasm detection and classification. In the first phase, the development of a mechanism for detecting sarcasm on bilingual data was explored. To achieve this, a feature extraction process was proposed to identify sarcasm features. Five feature categories of that can be extracted using natural language processing were considered: lexical, pragmatic, prosodic, syntactic and idiosyncratic. A non-linear Support Vector Machines classifier was employed to measure the performance of the features using the adopted evaluation metric, average Fmeasure. The best-performing features were then used as input for the second phase. In the second phase, a framework for sentiment analysis that considers sarcasm detection and classification was proposed. The framework consists of six modules, namely preprocessing, feature extraction, feature selection, sentiment classification, sarcasm detection and classification, and actual sentiment classification. Results obtained from the evaluation conducted demonstrate that the proposed features and framework are able to improve the performance of sentiment analysis. The best performance for sarcasm detection was found using a combination of syntactic, pragmatic, and prosodic features with an average F-measure score of 0.852. The best result of sentiment classification using the proposed framework, considering both sarcasm detection and classification, recorded an average F-measure score of 0. 905, outperforming the baseline sentiment classification score of 0.839. 2017 Thesis https://eprints.ums.edu.my/id/eprint/37665/ https://eprints.ums.edu.my/id/eprint/37665/1/24%20PAGES.pdf text en public https://eprints.ums.edu.my/id/eprint/37665/2/FULLTEXT.pdf text en validuser masters Universiti Malaysia Sabah Fakulti Komputeran dan Informatik
institution	Universiti Malaysia Sabah
collection	UMS Institutional Repository
language	English English
topic	HM711-806 Groups and organizations
spellingShingle	HM711-806 Groups and organizations Mohd Suhairi Md Suhaimin Sarcasm detection and classification to support sentiment analysis : A study in malay social media
description	The classification of users' sentiment from social media data can be used to determine public opinion on certain issues. The presence of sarcasm in text may hamper the performance of sentiment analysis. This thesis presents research work conducted on sarcasm detection and classification to support sentiment analysis. A Malay social media dataset, specifically focused on economic and political domain, was acquired from public comments posted on Facebook. The proposed work consists of two phases: (i) sarcasm detection and (ii) sentiment analysis with sarcasm detection and classification. In the first phase, the development of a mechanism for detecting sarcasm on bilingual data was explored. To achieve this, a feature extraction process was proposed to identify sarcasm features. Five feature categories of that can be extracted using natural language processing were considered: lexical, pragmatic, prosodic, syntactic and idiosyncratic. A non-linear Support Vector Machines classifier was employed to measure the performance of the features using the adopted evaluation metric, average Fmeasure. The best-performing features were then used as input for the second phase. In the second phase, a framework for sentiment analysis that considers sarcasm detection and classification was proposed. The framework consists of six modules, namely preprocessing, feature extraction, feature selection, sentiment classification, sarcasm detection and classification, and actual sentiment classification. Results obtained from the evaluation conducted demonstrate that the proposed features and framework are able to improve the performance of sentiment analysis. The best performance for sarcasm detection was found using a combination of syntactic, pragmatic, and prosodic features with an average F-measure score of 0.852. The best result of sentiment classification using the proposed framework, considering both sarcasm detection and classification, recorded an average F-measure score of 0. 905, outperforming the baseline sentiment classification score of 0.839.
format	Thesis
qualification_level	Master's degree
author	Mohd Suhairi Md Suhaimin
author_facet	Mohd Suhairi Md Suhaimin
author_sort	Mohd Suhairi Md Suhaimin
title	Sarcasm detection and classification to support sentiment analysis : A study in malay social media
title_short	Sarcasm detection and classification to support sentiment analysis : A study in malay social media
title_full	Sarcasm detection and classification to support sentiment analysis : A study in malay social media
title_fullStr	Sarcasm detection and classification to support sentiment analysis : A study in malay social media
title_full_unstemmed	Sarcasm detection and classification to support sentiment analysis : A study in malay social media
title_sort	sarcasm detection and classification to support sentiment analysis : a study in malay social media
granting_institution	Universiti Malaysia Sabah
granting_department	Fakulti Komputeran dan Informatik
publishDate	2017
url	https://eprints.ums.edu.my/id/eprint/37665/1/24%20PAGES.pdf https://eprints.ums.edu.my/id/eprint/37665/2/FULLTEXT.pdf
_version_	1794022677403402240

Sarcasm detection and classification to support sentiment analysis : A study in malay social media

Similar Items