有大量使用者產生的資料被儲存在網路上，而這些資料在許多研究領域上都是豐富的分析來源，例如選舉結果的預測、股市分析或是用於了解客戶滿意度，而情緒分類法亦可以應用在此。 此前 Argueta 提出了一個基於圖像的情緒分類法則可以克服這種挑戰。雖然這樣的方式能應用在大部分的西方語言上，但應用於其他尚未測試過的語言上可能有全然不同的結果。此外，儘管對於每一篇 Twitter 貼文可能隱含不同數量的情緒，但該系統仍回傳固定數量的情緒數量。
在此篇論文試著將此系統應用於日本語上並提出一個統計方法去調整情緒數量的結果，透過實驗去闡明情感強度與準確度之前的關係。我們的結果顯示該系統可以應用於日本語上，但當使用不同的字詞分割方式時，會有不同的精確度。透過此論文中所提出的方法來成功控制回傳情緒數量，進而改善了整體的精準度。結果表明，當情感的強度越強，精準度會更高。

Huge amount of user-generated data are stored in the web. They are rich resources for analysis in many areas such as election result prediction, stock market analysis, and knowing customer satisfaction. Emotion classification can be used for these analysis. A graph-based approach was proposed by Argueta which can overcome the challenges of emotion classification. Even though it works for major Western languages, the other languages which are totally different from them are not tested. Also, the system returns fixed number of emotion results even though the number of emotions inside a tweet are not same. In this paper, we adopted the system to Japanese languages and proposed a statistical method to change the number of emotion results. Besides, the experiment to clarify the relation between the intensity of emotion and accuracy is conducted. Our results show that the system works also for Japanese but the accuracy differs when applying different segmentation methods are applied. By our method to change the number of emotion results, overall accuracy was improved. The result shows that when the intensity of emotion is strong, the accuracy will be higher.

Abstract i
Acknowledgement ii
List of Figures v
1 Introduction 1
2 Related Work 4
2.1 Sentiment Analysis on Microblogs . . . . . . . . . . . . . . . . . . . . . . 4
2.2 Lexicon-based Approaches . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.3 Sentiment Analysis on Japanese Texts . . . . . . . . . . . . . . . . . . . . 6
2.4 Multilingual Approaches . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
3
Methodology
8
3.1 Data Collection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3.2 Preprocessing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.3 Pattern Extraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.4 Emotion Classification . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.5
4
Experiments
4.1
4.2
5
A Statistical Method to Select Emotions . . . . . . . . . . . . . . . . . . . 15
19
Experimental Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
4.1.1 Datasets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
4.1.2 Evaluation method . . . . . . . . . . . . . . . . . . . . . . . . . . 21
4.1.3 System Configuration . . . . . . . . . . . . . . . . . . . . . . . . 21
Experimental Result . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
4.2.1 Comparison of tokenizer and n-grams method . . . . . . . . . . . . 21
4.2.2 Evaluation of a statistical approach . . . . . . . . . . . . . . . . . 25
4.2.3 Evaluation on the intensity of emotion . . . . . . . . . . . . . . . . 28
Conclusions and Future Works
31
5.1 Conclustions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
5.2 Future Works . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
References 33

[1] Andreas Jungherr. Tweets and votes, a special relationship: the 2009 federal election
in germany. In Proceedings of the 2nd workshop on Politics, elections and data, pages
5–14. ACM, 2013.
[2] Manish Gaurav, Amit Srivastava, Anoop Kumar, and Scott Miller. Leveraging can-
didate popularity on twitter to predict election outcome. In Proceedings of the 7th
Workshop on Social Network Mining and Analysis, page 7. ACM, 2013.
[3] Erik Tjong Kim Sang and Johan Bos. Predicting the 2011 dutch senate election results
with twitter. In Proceedings of the Workshop on Semantic Analysis in Social Media,
pages 53–60. Association for Computational Linguistics, 2012.
[4] Nugroho Dwi Prasetyo and Claudia Hauff. Twitter-based election prediction in the
developing world. In Proceedings of the 26th ACM Conference on Hypertext & Social
Media, pages 149–158. ACM, 2015.
[5] Yuexin Mao, Wei Wei, and Bing Wang. Twitter volume spikes: analysis and applica-
tion in stock trading. In Proceedings of the 7th Workshop on Social Network Mining
and Analysis, page 4. ACM, 2013.
[6] Boram Park, Kibeom Lee, and Namjun Kang. The impact of influential leaders in
the formation and development of social networks. In Proceedings of the 6th Interna-
tional Conference on Communities and Technologies, pages 8–15. ACM, 2013.
[7] Akiyo Nadamoto, Mai Miyabe, and Eiji Aramaki. Analysis of microblog rumors and
correction texts for disaster situations. In Proceedings of International Conference
on Information Integration and Web-based Applications & Services, page 44. ACM,
2013.
[8] Carlos Argueta, Elvis Saravia, and Yi-Shin Chen. Unsupervised graph-based patterns
extraction for emotion classification. In Proceedings of the 2015 IEEE/ACM Interna-
tional Conference on Advances in Social Networks Analysis and Mining 2015, pages
336–341. ACM, 2015.
[9] Yuki Yamamoto, Tadahiko Kumamoto, and Akiyo Nadamoto. Role of emoticons
for multidimensional sentiment analysis of twitter. In Proceedings of the 16th In-
ternational Conference on Information Integration and Web-based Applications &
Services, pages 107–115. ACM, 2014.
[10] Fajri Koto and Mirna Adriani. Hbe: Hashtag-based emotion lexicons for twitter senti-
ment analysis. In Proceedings of the 7th Forum for Information Retrieval Evaluation,
pages 31–34. ACM, 2015.
[11] Robert Plutchik. The nature of emotions human emotions have deep evolutionary
roots, a fact that may explain their complexity and provide tools for clinical practice.
American Scientist, 89(4):344–350, 2001.
[12] Georgios Kalamatianos, Dimitrios Mallis, Symeon Symeonidis, and Avi Arampatzis.
Sentiment analysis of greek tweets and hashtags using a sentiment lexicon. In Pro-
ceedings of the 19th Panhellenic Conference on Informatics, pages 63–68. ACM,
2015.
[13] Felipe Bravo-Marquez, Eibe Frank, and Bernhard Pfahringer. From unlabelled tweets
to twitter-specific opinion words. In Proceedings of the 38th International ACM SI-
GIR Conference on Research and Development in Information Retrieval, pages 743–
746. ACM, 2015.
[14] Yoshimitsu Torii, Dipankar Das, Sivaji Bandyopadhyay, and Manabu Okumura. De-
veloping japanese wordnet affect for analyzing emotions. In Proceedings of the
2nd Workshop on Computational Approaches to Subjectivity and Sentiment Analy-
sis, pages 80–86. Association for Computational Linguistics, 2011.
[15] Michal Ptaszynski, Rafal Rzepka, Kenji Araki, and Yoshio Momouchi. Automatically
annotating a five-billion-word corpus of japanese blogs for affect and sentiment anal-
ysis. In Proceedings of the 3rd Workshop in Computational Approaches to Subjectiv-
ity and Sentiment Analysis, pages 89–98. Association for Computational Linguistics,
2012.
[16] Xiaojun Wan. Co-training for cross-lingual sentiment classification. In Proceedings of
the Joint Conference of the 47th Annual Meeting of the ACL and the 4th International
Joint Conference on Natural Language Processing of the AFNLP: Volume 1-Volume
1, pages 235–243. Association for Computational Linguistics, 2009.
[17] Valentin Jijkoun and Katja Hofmann. Generating a non-english subjectivity lexicon:
relations that matter. In Proceedings of the 12th Conference of the European Chapter
of the Association for Computational Linguistics, pages 398–405. Association for
Computational Linguistics, 2009.