ERA

Download the full-sized PDF of Emotion Mining from TextDownload the full-sized PDF

Analytics

Share

Permanent link (DOI): https://doi.org/10.7939/R3C53F63N

Download

Export to: EndNote  |  Zotero  |  Mendeley

Communities

This file is in the following communities:

Graduate Studies and Research, Faculty of

Collections

This file is in the following collections:

Theses and Dissertations

Emotion Mining from Text Open Access

Descriptions

Other title
Subject/Keyword
emotion detection
text mining
sentiment analysis
Type of item
Thesis
Degree grantor
University of Alberta
Author or creator
Gholipour Shahraki, Ameneh
Supervisor and department
Zaiane, Osmar (Computing Science)
Examining committee member and department
Stroulia, Eleni (Computing Science)
Elmallah, Ehab (Computing Science)
Department
Department of Computing Science
Specialization

Date accepted
2015-09-25T15:23:56Z
Graduation date
2015-11
Degree
Master of Science
Degree level
Master's
Abstract
Emotion mining is the science of detecting, analyzing, and evaluating humans’ feelings towards different events, issues, services, or any other interest. One of its specific directions is text emotion mining, that refers to analyzing people’s emotions based on observations of their writings. Text emotion mining is an interdisciplinary topic of interest and has many applications including helping customer care services, recommending music or movies to computer users, helping in selecting e-learning materials, filtering results of searches by emotion, and diagnosing depression or suicidal tendency. In this work, we study the problem of text emotion classification. First, we collect and cleanse a corpus of Twitter messages that convey at least one of the emotions: anger, fear, joy, love, sadness, surprise, disgust, guilt, and thankfulness. Then, we propose several lexical and learning based methods to classify the emotion of test tweets and study the effect of different feature sets, dimension reduction techniques, different learning algorithms and configurations, and also try to address the problem of sparsity of the input data. Our experimental results show that a set of Naive Bayes classifiers, each corresponding to one emotion, using unigrams as features, is the best-performing method for the task. Moreover, we address the problem of multi-label emotion classification of texts, that is concerned with tweets that represent more than one emotion. In this case, again the Naive Bayes method outperforms the others. In order to compare the efficiency of our algorithms, we test them also on a couple of other datasets, one of which is collected from Twitter, and the other contains a set of formally written texts. Our Naive Bayes approach achieves higher accuracy, compared with state-of-the-art methods working on these corpora.
Language
English
DOI
doi:10.7939/R3C53F63N
Rights
Permission is hereby granted to the University of Alberta Libraries to reproduce single copies of this thesis and to lend or sell such copies for private, scholarly or scientific research purposes only. The author reserves all other publication and other rights in association with the copyright in the thesis and, except as herein before provided, neither the thesis nor any substantial portion thereof may be printed or otherwise reproduced in any material form whatsoever without the author's prior written permission.
Citation for previous publication

File Details

Date Uploaded
Date Modified
2015-09-25T21:23:57.491+00:00
Audit Status
Audits have not yet been run on this file.
Characterization
File format: pdf (PDF/A)
Mime type: application/pdf
File size: 759892
Last modified: 2016:06:24 18:35:15-06:00
Filename: Gholipour Shahraki_Ameneh_201509_MSc.pdf
Original checksum: 174b16acc0e83617d5fcb8a52b0862f0
Well formed: true
Valid: true
Page count: 72
Activity of users you follow
User Activity Date