臺灣博碩士論文加值系統

情感分析(Sentiment Analysis)是近年來在資料探勘中相當熱門的研究議題，
本研究提出一個新的情感分析機制。此機制結合過去主流的字典法(Lexicon
Based Methods)與機器學習法(Machine Learning Based Methods)，以情感辭典作為
情感遮罩有效減少機器學習法的特徵，並加入以字典法為基礎所計算出五個情感
值指標，作為新特徵加入到機器學習的模型中進行訓練。本研究根據情感遮罩和
情感值指標的不同組合，共提出了 15 種的情感分析方法，並通過貝氏分類器
(naive Bayes classifiers)、支援向量機(SVM)，和兩種不同的語料庫進行實驗。經
由實驗結果本研究發現貝氏分類器的性能優於支援向量機，且經由情感遮罩過濾
特徵後的實驗結果，其準確率並無下降，但訓練時程得到明顯的優化。而通過決
策樹(Decision tree)的分析，本研究發現如果在特徵中包含正向詞詞頻、負向詞詞
頻、正向詞和負向詞詞頻差值這 3 個情感值指標，能更有效的提升情感分析的準
確率。

Sentiment Analysis is a very popular research topic in data mining in recent years.This study proposes a new sentiment analysis mechanism. This mechanism combines the Lexicon Based Methods and Machine Learning Based Methods and uses the sentiment dictionary as sentiment masks to effectively reduce the features of machine learning, and adds five sentiment indicators based on the Lexicon Based Methods. The sentiment indicators are added to the machine learning model as new features for training. Based on the different combinations of sentiment masks and sentiment indicators, this study proposes 15 kinds of sentiment analysis methods, which are carried out experiments by naive Bayes classifiers, support vector machines (SVM),and two different corpora. Through the experimental results, this study found that the performance of the naive Bayes classifiers is better than support vector machine, and the accuracy of the experimental results after filtering the features through the sentiment mask does not decrease, but the training time is significantly optimized. Through the analysis of the decision tree, this study finds that if the feature includes this three sentiment indicators: positive word frequency, negative word frequency, positive word and negative word frequency difference, it can improve the accuracy of sentiment analysis more effectively.

審定書............................................................................................................................ I
致謝...............................................................................................................................II
摘要..............................................................................................................................III
Abstract........................................................................................................................IV
Contents........................................................................................................................V
List of figures...............................................................................................................VI
List of tables............................................................................................................... VII
Chapter 1. Introduction ..................................................................................................1
1.1 Background and motivation...............................................................................................................1
1.2 Research purpose...............................................................................................................................3
1.3 Research process ...............................................................................................................................3
Chapter 2. Iiterature review ...........................................................................................6
2.1 Sentiment analysis.............................................................................................................................6
2.2 Lexicon based methods for sentiment analysis..................................................................................7
2.3 Machine learning based methods for sentiment analysis...................................................................9
2.4 Chinese word segmentation.............................................................................................................10
Chapter 3. New sentence analysis framework basic on lexicon based method and
machine learning method.............................................................................................11
3.1 Data sets description........................................................................................................................11
3.2 Word segmentation and feature selection ........................................................................................12
3.3 Building a sentiment dictionary and designing two masks for feature filtering ..............................13
3.4 Combination design of different masks and lexicon based features................................................18
3.5 Extend the sentiment dictionary with Word2vec .............................................................................20
3.6 New sentiment analysis mechanism using the sentiment feature combinations..............................21
Chapter 4. Experimental results...................................................................................22
4.1 The first experiment ........................................................................................................................22
4.2 The second experiment....................................................................................................................24
4.3 The third experiment .......................................................................................................................35
4.4 Negation words as negative words..................................................................................................39
Chapter 5. Conclusions................................................................................................41
Reference .....................................................................................................................43

中文文獻
[1] 林振穎, 2017, 從新聞文章預測股票走勢:使用粒子群演算法與情緒分析, 國
立高雄應用科技大學資訊管理研究所碩士論文.
[2] 孫勁光, 馬志芳, 孟祥福, 2013, 基於情感詞屬性和雲模型的文本情感分類
方法,電腦工程 Vol. 39, No. 12 ,pp. 211-215.
[3] 高紹航, 高照明, 2007, 詞義辨識:機器學習演算法特徵的選取與組合,
Proceedings of the 19th Conference on Computational Linguistics and Speech
Processing, pp. 131-144.
[4] 陳安怡, 2017, 運用文字探勘及情緒分析技術發展店家品項評價模組, 國立
台北大學資訊管理研究所碩士論文.
[5] 陳瑤光, 賴聯福, 黃亮綜, 伍朝欽, 2017, 使用情緒分析來探勘社群網路中的
負向個案, 2017健康與管理學術研討會, 論文編號:05-19-22-58-41-44, pp. 1-8.
[6] 曾婉菁, 2018, 機器學習探究, 中華印刷科技學會, Vol. 34, No. 2, pp. 1-32.
[7] 楊立公, 朱儉, 湯世平, 2013,文本情感分析綜述, 電腦應用, Vol. 33, No. 6, pp.
1574-1607.
[8] 楊惠淳, 2011, 以主客觀分析與相互資訊檢索探討情感分析之準確度-以電影
評論為例, 國立臺北科技大學資訊與運籌管理研究所碩士論文.
[9] 趙鵬, 趙志偉, 卓景文, 2011, 一種情感詞語義加權的句子傾向性識別方法,
電腦工程與應用, Vol. 47, No. 35, pp. 161-163.
[10] 劉佳欣, 2018, 以社群網路之情緒分析預測汽車銷售市場, 國立暨南國際大
學資訊管理研究所碩士論文.
[11] 酆偉寬, 2015, 網路情感分析對於手機應用程式評價之影響的研究, 淡江大
學資訊管理學系研究所碩士論文.
英文文獻
[12] Nur Azizah Vidya, Mohamad Ivan Fanany and Indra Budi, 2015, Twitter sentiment
to analyze net brand reputation of mobile phone providers, Procedia Computer
Science, Vol. 72, pp. 519-526.
[13] Barkha Bansal and Sangeet Srivastava, 2018, On predicting elections with hybrid
topic based sentiment analysis of tweets, Procedia Computer Science, Vol. 135,
pp. 346-353.
[14] Bernhard Boser, Isabelle Guyon and Vladimir Vapnik, 1992, A training algorithm
for optimal margin classifiers, COLT '92 Proceedings of the fifth annual workshop
on Computational learning theory, pp. 144-152.
44
[15] Marouane Birjali, et al., 2017, Machine Learning and Semantic Sentiment
Analysis based Algorithms for Suicide Sentiment Prediction in Social Networks,
Procedia Computer Science, Vol.113, pp. 65-72.
[16] Muhammad Bilal, et al., 2016, Sentiment classification of Roman-Urdu opinions
using Naïve Bayesian, Decision Tree and KNN classification techniques, Journal
of King Saud University - Computer and Information Sciences, Vol. 28, No. 3, pp.
330-344.
[17] Robert Baumgartner, Wolfgang Gatterbauer and Georg Gottlob, 2009,
Encyclopedia of Database Systems, New York, Springer-Verlag.
[18] Corinna Cortes and Vladimir Vapnik, 1995, Support-vector networks, Machine
learning, Vol. 20, No. 3, pp. 273-297.
[19] Li Dong, et al., 2014, Adaptive recursive neural network for target-dependent
twitter sentiment classification, Proceedings of the 52nd Annual Meeting of the
Association for Computational Linguistics, Vol. 2, pp. 49-54.
[20] Namrata Godbole, Manjunath Srinivasaiah and Steven Skiena, 2007, Large-Scale
Sentiment Analysis for News and Blogs, The International Conference on Weblogs
and Social Media. (ICWSM), Vol. 7, No. 21, pp. 219-222.
[21] Minqing Hu, and Bing Liu, 2004, Mining and summarizing customer reviews,
Proceedings of the tenth ACM SIGKDD international conference on Knowledge
discovery and data mining, pp. 168-177.
[22] Minqing Hu, and Bing Liu, 2004, Mining opinion features in customer reviews,
Proceedings of the 19th National Conference on Artificial Intelli2 gence: AAAI
2004, Vol. 4, No. 4, pp. 755-760.
[23] Stuart Jonathan Russell and Peter Norvig, 2002, Artificial Intelligenz-A Modern
Approach, 2nd, New Jersey, Prentice Hall.
[24] Yang Li, et al., 2016, A web sentiment analysis method on fuzzy clustering for
mobile social media users, EURASIP Journal on Wireless Communications and
Networking, Vol. 2016, No. 1, pp. 128.
[25] Tetsuya Nasukawa, and Jeonghee Yi, 2003, Sentiment analysis: Capturing
favorability using natural language processing, Proceedings of the 2nd
international conference on Knowledge capture, pp. 70-77.
[26] Bo Pang and Lillian Lee, 2008, Opinion mining and sentiment analysis,
Foundations and Trends in Information Retrieval, Vol. 2, No. 1-2, pp. 1-135.
[27] Jin-Shu Su, Bo-Feng Zhang and Xin Xu, 2006, Advances in machine learning
based text categorization, Journal of Software, Vol. 17, No. 9, pp. 1848-1859.
[28] Yang Shen, Shuchen Li, Ling Zheng, Xiaodong Ren, and Xiaolong Cheng, 2009,
Emotion mining research on micro-blog, 2009 1st IEEE symposium on web society,
pp. 71-75.
45
[29] Ying-Tse Sun, Chien-Liang Chen, Chun-Chieh Liu, Chao-Lin Liu, and Von-Wun
Soo, 2010, Sentiment Classification of Short Chinese Sentences, Proceedings of
the 22nd Conference on Computational Linguistics and Speech Processing
(ROCLING 2010), pp. 184-198.
[30] Duyu Tang, et al., 2014, Coooolll: A deep learning system for twitter sentiment
classification, Proceedings of the 8th international workshop on semantic
evaluation, pp. 208–212.
[31] Mike Thelwall, et al., 2010, Sentiment strength detection in short informal text,
Journal of the American Society for Information Science and Technology, Vol. 61,
No. 12, pp. 2544-2558.
[32] Chung-Hsien Wu, Ze-Jing Chuang, and Yu-Chung Lin, 2006, Emotion
recognition from text using semantic labels and separable mixture models, ACM
transactions on Asian language information processing (TALIP) Vol. 5, No. 2,
pp. 165-183.
[33] Fredric Wolf, 1986, Meta-analysis: Quantitative methods for research synthesis,
Thousand Oaks, Sage Publications.
[34] Suchita Wawre and Sachin Deshmukh, 2016, Sentiment classification using
machine learning techniques, International Journal of Science and Research
(IJSR), Vol.5, No. 4, pp. 819-821.
[35] Lei Zhang, Shuai Wang and Bing Liu, 2018, Deep learning for sentiment analysis:
A survey, Wiley Interdisciplinary Reviews: Data Mining and Knowledge
Discovery, Vol. 8, No. 4, pp. 1-25.
[36] Lin Zhang, et al., 2014, Sentiment analysis on reviews of mobile users, Procedia
Computer Science, Vol. 34, pp. 458-465.
網路資料
[37] mathlf2015, 2016, Tmall Mall Sentiment Dictionary, from
https://github.com/mathlf2015/text_analysis.
[38] Natural Language Processing Laboratory, 2015, National Taiwan University
Sentiment Dictionary (NTUSD), from
http://nlg.csie.ntu.edu.tw/groupmember.php.
[39] YZU NLP Lab, 2016, Chinese Valence-Arousal Text (CVAT), from
http://nlp.innobic.yzu.edu.tw/resources/cvat.html.
[40] YZU NLP Lab, 2019, The Chinese Valence-Arousal Words (CVAW), from
http://nlp.innobic.yzu.edu.tw/resources/cvaw.html.
[41] BUPTLdy, 2016, Chinese Shopping Reviews, from
https://github.com/BUPTLdy/Sentiment-Analysis