臺灣博碩士論文加值系統

「科技始終來自於人性」，而情緒是人類表達的基礎，在目前的應用軟體中，文字扮演著一個重要的角色，藉由文字來抒發情緒。所以，讓電腦從文字當中具備察覺和回應人類情緒的能力，可以使人機互動更加自然。根據上述的需求，本論文在首先提出了一個具有高辨識能力、不需要大量使用資料庫，從事件層級文字來自動偵測主角人物(Subject)的情緒。在系統中，預先挑選數組有代表性的主、受詞組合加上預先標注該組合與各種動作之間的可能情緒反應，在測試的文字中分析該句中的主詞與受詞間之相互行為來配對預設配對中具有最相似的相互行為。其中結合網頁資料挖礦(web mining)及語意角色分析(semantic role labeling)的技術以偵測出文字中主角人物是正面、負面或是中性情緒。藉由實驗結果得知，所提出的方法能達到接近85%的準確率。
在論文的第二部份，則使用了機器學習方法來偵測多重的情緒。我們使用了三種分類器來進行這個任務，這三種分類器包含 “最近鄰居搜尋”, “支持向量機”, 以及“決策樹演算法”。我們將大量的事件層級文字用人工方法標上情緒後，再運用上述之網頁資料挖礦(web mining)及語意角色分析(semantic role labeling)的技術來得到該句的相互行為分析圖以做為機器學習的輸入。藉由實驗來比較出這三種分類器的效能，決策樹演算法表現出將近70%正確率來偵測由[Ekman 1993] 所提出的六種基本情緒，高興、難過、生氣、憤怒、驚喜、噁心以及無情緒。
在論文的最後，我們實作一個以AIML為標準架構的網路聊天機器人，其特色除了有大量的常識對話外，另結合了上述的文字情緒偵測機制，文字偵測機制的加入不但能讓使用者與機器人聊天時增加娛樂性，機器人更加深入地理解訊息中的情緒並且應答出更貼切的回答。實驗結果也展現出結合了文字情緒偵測功能的聊天機器人更能增加使用時的娛樂性、功能性，使用者也正面表示願意再次使用此系統。

Automatic emotion sensing in textual data is crucial for the development of intelligent interfaces in many interactive computer applications. This thesis first describes a high-precision, knowledgebase-independent approach for automatic emotion sensing for the subjects of events embedded within sentences. The proposed approach is based on the probability distribution of common mutual actions between the subject and the object of an event. We have incorporated web-based text mining and semantic role labeling techniques, together with a number of reference entity pairs and hand-crafted emotion generation rules to realize an event emotion detection system. The evaluation outcome reveals a satisfactory result with about 85% accuracy for detecting the positive, negative and neutral emotions.

In the second part of this thesis, machine learning models are applied to predict multiple emotions from event-level sentences. We have conducted experiments using three machine learning models, namely, KNN, SVM, and Decision Tree algorithm. By adopting the MAH as the major feature for the machine learning model, the best precision achieved about 70% using decision tree algorithm for detecting Big-Six emotions [Ekman 1993] including “Happy”, “Sad”, “Fear”, “Surprised”, “Disgusted”, “Angry” and “Neutral”.

Finally, a system framework of an emotion-aware chat bot system was proposed. The distinct features of our chat bot is able to express sentences that appropriately response to the emotion status of the user uttering the sentences. A prototype system has incorporated an emotion engine developed based on machine learning model to detect Big-Six emotions including “Happy”, “Sad”, “Fear”, “Surprised”, “Disgusted”, “Angry” and “Neutral”. A number of example runs are given to demonstrate the capabilities of the system that shows the applicability of proposed methodology in real-life applications.

Acknowledgements v
Contents vi
List of Figures ix
List of Tables x
Chapter 1. Introduction - 1 -
1.1. Emotion Sensing Incorporated Applications - 1 -
1.2. Motivations - 2 -
1.3. Problem Definition - 3 -
1.4. Contributions and Thesis Architecture - 4 -
Chapter 2. Related Research 6
2.1. Keyword based Approach 6
2.2. Analysis by Lexical Affinity 7
2.3. Analysis by Commonsense Knowledge 8
2.4. Analysis by Semantic Labels (Textual Definitions) 10
2.5. Summary 12
Chapter 3. Event-Level Emotion Sensing Based on Mutual Actions Between Entities 15
3.1. Overall of Emotion Sensing based on Mutual Actions 15
3.2. Semantic Role Labeling 17
3.3. Web Scale Text Mining for Mutual Actions 19
3.3.1. Verb categorization 21
3.4. Reference Entity Pairs (RE-Pairs) with Emotion Generation Rules 24
3.5. Matching MAHs among RE-Pairs 25
3.6. Experiments and Results 26
3.6.1. The Training Phase: 27
3.6.2. The Testing Phase: 28
3.6.3. Experiment Results: 29
3.7. Observation 30
3.7.1. Context-sensitive problems vs. Emotion sensing 32
Chapter 4. Applying Machine Learning Models for Multiple Emotion Sensing 34
4.1. EGRs of RE-Pairs for Multiple Emotion Detection 34
4.1.1. Reference Entity Pairs (RE-Pairs) with Emotion Generation Rules (EGRs) based on Multiple Emotions (Big-Six emotion [Ekman 1993]) Annotation 35
4.1.2. Matching MAHs among RE-Pairs 36
4.1.3. Experiment and Results 37
4.1.4. Observations 39
4.2. Machine Learning Model 41
4.3. Dataset Collection 43
4.4. Feature Extraction and Format 45
4.5. Classifiers 47
4.5.1. K-Nearest Neighborhood Algorithm 47
4.5.2. Support Vector Machine 49
4.5.3. Decision Tree Algorithm 50
4.6. Experiment Results 51
4.6.1. Training Phase and Testing Phase 51
4.6.2. Experiment Results 52
4.7. Observations 57
4.8. Model Refinement 58
4.8.1. MAH style selection 58
4.9. Discussion 60
Chapter 5. An Emotion-aware Chat Bot System 62
5.1. System Architecture 62
5.2. Hybrid Emotion Engine 63
5.3. Flash-based Animation Generator 64
5.4. Alice Chat Bot Module: A Simple Primer 65
5.4.1. Category Definition 66
5.4.2. Symbolic Reduction 67
5.4.3. Synonym Processing 68
5.4.4. Keywords Processing 69
5.5. Chat Bot Knowledgebase 71
5.6. User Study and Satisfactory Evaluation 74
5.7. Summary 76
Chapter 6. Conclusions and Future Work 78
Reference 80
Appendix1: Synonym groups for the verbs 85
Appendix 2: Emotion rules for the reference entity pair with “she” as the subject and “diamond ring” as the object. 87
Appendix 3: A list of tested events with the majority vote n > 9. 88
Appendix 4: A sample AIML document generated by the event-level emotion engine. 89

[Alice 2002] A.L.I.C.E. Artificial Intelligenence Foundation 2002. http://www.alicebot.org
[Augello et al. 2003] A. Augello, G. Saccone, S. Gaglio, and G. Pilato, “Humorist Bot: Bring Computational Humor in a Chat-Bot System,” in Proceedings of International Conference on Complex, Intelligent and Software Intensive Systems 2008, pp.703-708.
[Bartneck 2002] C. Bartneck, “Integrating the OCC Model of Emotions in Embodied Characters,” in Proceedings of the Workshop on Virtual Conversational Characters: Applications, Methods, and Research Challenges, 2002.
[Bates 1994] J. Bates, “The role of emotion in believable agents,” Communications of the ACM, vol. 37, no. 7, pp. 122-125, 1994.
[C4.5 1993] J.R. Quilan, “C4.5: Programs for Machine Learning,” Morgan Kaufmann Publishers, 1993
[Carreras and Marquez 2005] X. Carreras and L. Marquez, “Introduction to the CoNLL-2005 Shared Task: Semantic Role Labeling,” in Proceedings of Conference on Computational Natural Language Learning, 2005.
[Dasarathy 1991] B.V. Dasarathy (Editor), “Nearest Neighbor Norms: NN Patterns Classification Techniques” ISBN: 0-8186-8930-7
[Dyer 1987] M. Dyer, “Emotions and Their Computations: Three Computer Models,” Cognition and Emotion, vol. 1, no. 3, pp. 323-347, 1987.
[Dang and Palmer 2005] H. Dang and M. Palmer, “The role of semantic roles in disambiguating verb senses,” in Proceedings of Association for Computational Linguistics-05, 2005.
[Elliott 1992] C. Elliott, “The Affective Reasoner: A Process Model of Emotions in a Multi-agent System,” Ph.D. dissertation, Northwestern University, The Institute for the Learning Sciences, Technical Report No. 32, 1992.
[Etzioni et al. 2005] O. Etzioni, M. Cafarella, and D. Downey, A. Popescu, T. Shaked, S. Soderland, D. S. Weld, and A. Yates, “Unsupervised named-entity extraction from the Web: an experimental study,” Artificial Intelligence, vol. 165, pp. 91-134, 2005.
[Ekman 1993] P. Ekman, “Facial Expression of Emotion,” American Psychologist, vol.48, pp. 384-392, 1993.
[Fellbaum 1998] C. Fellbaum , WordNet: A Lexical Database. The MIT Press, 1998.
[Freese 2007] E. Freese, “Enhancing AIML Bots using Semantic Web Technologies,” In Proceedings of Extreme Markup Languages, 2007.
[Girju et al. 2006] R. Girju, A. Badulescu, and D. Moldovan, “Automatic discovery of part-whole relations,” Computational Linguistics, vol. 32, no. 1, pp. 83-135, 2006.
[Gildea and Jurafsky 2002] D. Gildea and D. Jurafsky, “Automatic Labeling of Semantic Roles,” Computational Linguistics, vol.28, no. 3, pp. 245-288, 2002.
[Hearst 1992] M. Hearst, “Automatic Acquisition of Hyponyms from Large Text Corpora,” in Proceedings of COLING, 1992.
[Koomen et al. 2005] P. Koomen, V. Punyakanok, D. Roth, and W. Yih, “Generalized Inference with Multiple Semantic Role Labeling System,” in Proceedings of Conference on Computational Natural Language Learning, 2005.
[Kullback and Leibler 1951] S. Kullback and R.A. Leibler, “On information and sufficiency,” Annals of Mathematical Statistics, vol. 22, pp. 79-86, 1951.
[LIBSVM 2001] C. Chang and C. Lin, “LIBSVM : a library for support vector machines”, 2001. Software available at http://www.csie.ntu.edu.tw/~cjlin/libsvm
[Liu et al. 2003] H. Liu, H. Lieberman, and T. Selker, “A Model of Textual Affect Sensing using Real-World Knowledge,” in Proceedings of the 2003 International Conference on Intelligent User Interfaces, 2003, pp. 125-132.
[Lu et al. 2009] C-Y Lu, S-H Lin, J-C Liu, S. Cruz-Lara., and J-S Hong, “Automatic Event-Level Textual Emotion Sensing Using Mutual Action Histogram between Entities,” Expert Systems with Applications (2009), doi:10.1016/j.eswa.2009.06.099
[Nelson and Mateas 2007] M. Nelson and M. Mateas, “Towards Automated Game Design,” in Proceedings of artificial intelligence and human-oriented computing, 2007, pp. 626-637.
[Ortony et al. 1988] A. Ortony, G. Clore, and A. Collins, The cognitive structure of emotions. New York: Cambridge University Press, 1988.
[OpenMind Project] OpenMind Project: http://commons.media.mit.edu/en/
[Picard 1997] R.W. Picard, Affective Computing, The MIT Press, Mass., 1997.
[Pantel and Pennacchiotti 2006] P. Pantel and M. Pennacchiotti, “Espresso: Leveraging generic patterns for automatically harvesting semantic relations,” in Proceedings of the 21st Int. Conference on Computational Linguistics and the 44th annual meeting of the Association for Computational Linguistics, 2006, pp. 113-120.
[Punyakanok et al. 2005] V. Punyakanok, D. Roth, and W. Yih, “The Necessity of Syntactic Parsing for Semantic Role Labeling,” in Proceedings of the International Joint Conference on Artificial Intelligence, 2005, pp. 1117-1123.
[Pradhan et al. 2003] S. Pradhan, K. Hacioqlu, W. Ward, J.H. Martin, and D. Jurafsky, “Semantic role parsing: adding semantic structure to unstructure text,” in IEEE International Conference on Data Mining, 2003, pp. 629-632.
[Pradhan et al. 2004] S. Pradhan, W. Hacioglu, J. Martin, and D. Jurafsky, “Shallow Semantic Parsing using Support Vector Machines,” in Proceedings of the Human Language Technology Conference/North American chapter of the Association for Computational Linguistics annual meeting, 2004.
[Quinlan 1993] J.R. Quinlan, “C4.5: Programs for Machine Learning,” Morgan Kaufgmann Publishers, 1993.
[Rubner et al. 2000] Y. Rubner, C. Tomasi, and L. J. Guibas, “The earth mover's distance as a metric for image retrieval,”. International Journal of Computer Vision, 40(2):99-121, November 2000.
[Rzepka et al. 2009] R. Rzepka, W. Shi, M. Ptazzynski, P. Dybala, S. Higuchi, and K. Araki, “Serious Processing for Frivolous Purpose- A Chatbot Using Web-mining Supported Affect Analysis and Pun Generation”, in Proceedings of the 13th international conference on Intelligent user interfaces 2009, pp.487-488.
[Shaikh et al. 2007] Mostafa Shaikh, Helmut Prendinger, and Mitsuru Ishizuka, “SenseNet: A Linguistic Tool to Visualize Numerical-Valence based Sentiment of Textual Data,” (poster paper) Proc 5th Int'l Conf on Natural Language Processing (ICON-07), Hyderabad, India, Jan. 2007, pp 147-152.
[Shawar et al. 2003] A. Shawar, Bayan, and E. Atwell, “Using Dialogue Corpora to Retain a Chatbot System,” in Proceeding of the Corpus Linguistics 2003, pp.283-294.
[Strappa et al 2007] C. Strapparava, A. Valitutti, and O. Stock, “The Affective Weight of Lexicon”, in Proceedings of the Fifth International Conference on Language Resource and Evaluation, 2006.
[Soong et al. 2001] W.M. Soong, H.T. Ng, and D.C.Y. Lim, “A Machine Learning Approach to Coreference Resolution of Noun Phrases,” Computational Linguistics, vol. 27, no. 4, pp. 521-544, 2001.
[Tang et al. 2009] H. Tang, S. Tan, X. Cheng, “A Survey on Sentimental Detection of Reviews,” Expert Systems with Applications, vol. 36, issue 7, p.p. 10760-10773, 2009.
[Tatai et al. 2003] G. Tatai, A. Csordas, A. Szalo, and L. Laufer, “The Chatbot Feeling – Towards Animated Emotional ECAs”, in The 11th Portuduese Conference on Artificial Intelligence 2003, pp.336-340.
[Wilks 1999] Y. Wilks, “Machine Conversations,” Academic Pulisher, DordrechtKluwer, 1999.
[WordNet 1999] Christiane Fellbaum (Ed). 1999. WordNet: An Electronic Lexical Databases, MIT Press, Cambridge, Massachusetts
[WordNet-Affect 2004] C. Strapparava and A. Valitutti., “Wordnet-affect: An affective extension of wordnet,” in Proceedings of the 4th International Conference on Language Re-sources and Evaluation, Lisbon, 2004
[Wu et al. 2003] S.-H. Wu, T.-H. Tsai, and W.-L. Hsu, “Domain Event Extraction and Representation with Domain Ontology,” Proceedings of the IJCAI-03 Workshop on Information Integration on the Web, Acapulco, Mexico, 2003. pp. 33-38.
[Wu et al. 2006] C. Wu, Z. Chuang, and Y. Lin, “Emotion recognition from text using semantic labels and separable mixture models,” ACM Transactions on Asian Language Information Processing, vol. 5, no. 2, pp. 165-183, 2006.
[Yahoo Dictionary] Yahoo Online Dictionary, accesible at: http://dictionary.yahoo.com/
[Zhou and Huang 2003] X. Zhou and T. Huang, “Relevance feedback in image retrieval:A comprehensive review”, Multimedia System, vol 8, no. 6, pp. 536-544, 2003.