Sentimento - Opinion Mining System for Product Review / Emad Ashraf Samuel

Supervisor: Samhaa El-Beltagy

Thesis (M.A.)—Nile University, Egypt, 2015 .

"Includes bibliographical references"

Contents:
CHAPTER 1: INTRODUCTION 1
1.1. Motivation ............................................................................................2
1.2. Background ..........................................................................................3
1.3. Problem Definition..............................................................................5
1.4. Proposed Solution ...............................................................................7
1.5. Methodology ........................................................................................8
1.6. Thesis Outline ....................................................................................11
CHAPTER 2: LITERATURE REVIEW 13
2.1 About Sentiment Analysis ...............................................................14
2.2 Aspect-Based Sentiment Analysis Research .................................22
2.3 Our Approach in Sentimento ..........................................................32
CHAPTER 3: SYSTEM ARCHITECTURE 35
3.1 What is Sentimento? .........................................................................36
3.2 Sentimento’s Architecture ...............................................................37
3.3 Sentimento’s Stages ..........................................................................38
CHAPTER 4: ASPECTS EXTRACTION 45
4.1 Aspect-Based Ontology ....................................................................46
4.2 Aspects Extraction and Categorization in Sentimento ................52
CHAPTER 5: SENTIMENT CLASSIFICATION 55
5.1 Opinion Lexicon ................................................................................57
5.2 Sentiment Classification in Sentimento .........................................60
iv
CHAPTER 6: EVALUATION 65
6.1 Choosing Datasets For Evaluation .................................................67
6.2 SemEval Datasets ..............................................................................69
6.3 Evaluation of Aspects Extraction ....................................................71
6.4 Evaluation of Sentiment Classification ..........................................73
6.5 Evaluation of Other Parameters......................................................76
CHAPTER 7: CONCLUSION 79
7.1 Current Status of Sentimento ..........................................................80
7.2 Future Work .......................................................................................81
APPENDIX A – HOW TO USE SENTIMENTO 85
APPENDIX B – SAMPLES FROM EVALUATION DATASET 91
REFERENCES

Abstract:
and more and more people are writing online reviews. These reviews are very important to give the real experience of products to potential customers and to the producing corporations. The number of reviews for some products may reach hundreds or even thousands. We need an automatic system that is able to summarize these reviews in a way that shows what exactly people like and dislike about the product. In this work, we develop Sentimento as an opinion mining system for online product reviews that is able to provide an aspect-based summary with accuracy close to an expert human reviewer and within an acceptable time. Sentimento is mainly divided to two tasks: aspects extraction and categorization, and opinions extraction and classification. The main idea is using ontology for the first task and opinion lexicon for the second task. Our experimental results using annotated test data on laptop reviews are promising for both tasks.

Text in English, abstracts in English.