傳統文本情緒分類方法通常將文本分類至一個事先指定的離散情緒集合中的某一項情緒。但這類情緒分類的方法常常引起分類錯誤的問題以及用戶難以爲文本標註某一具體情緒的困難。有研究者提出了一種連續的情緒模型，將情緒定義在一個二維的連續空間內。在該連續空間內，標註這對該文字標註，以確定該段文字的情緒在該連續空間內的位置。但因標註者對文本情緒的解讀不一致，會導致情緒位置標註分散的問題。本研究提出了一種將傳統分類方法整合至連續情緒空間的概率模型。該模型中用valence（效價）和arousal（喚醒度）兩個維度定義一段文字的情緒。該概率模型可以同時避免傳統分類方法容易引起分類錯誤的問題和連續空間情緒模型中標註者標註分散的問題。本研究提出的概率模型不僅可以確定一段文本的情緒，同時還可以給出該情緒在二維連續空間中的分佈與趨勢。

Previous work on text emotion detection mainly aims at extracting an emotion of a given statement. In these work emotions are defined in a discrete set and the misclassification rate relies on the accuracy of specific classifier. Researchers also propose a continuous emotions model that an emotion is defined in a two-demotion continuous space. To determine emotions of a given statement in this space, annotators are asked to rate the statement, which easily causes divergence rating problem. This research proposes a probabilistic model that integrates determinative emotion classifiers into a continuous emotion model, in which an emotion is defined by valence and arousal, to predict the emotion distribution of a given statement. This predictive model is able to reduce the influence misclassification rate of emotion classifier and divergence rating caused by annotators at the same time. In addition, this model not only extracts a single emotion but also provides the potential emotions and emotion tendency with strength, which means, for a given text, quantitative relationships among possible emotions are available for further analysis, such as emotion strength detection and emotion changing over time.

1] A. Agarwal, B. Xie, I. Vovsha, O. Rambow, and R. Passonneau. Sentiment analy-sis of twitter data. InProceedings of the workshop on languages in social media,pages 30–38. Association for Computational Linguistics, 2011.
[2] C. Akkaya, A. Conrad, J. Wiebe, and R. Mihalcea. Amazon mechanical turk forsubjectivity word sense disambiguation. InProceedings of the NAACL HLT 2010Workshop on Creating Speech and Language Data with Amazon’s MechanicalTurk, pages 195–203. Association for Computational Linguistics, 2010.
[3] C. Argueta and Y.-S. Chen. Multi-lingual sentiment analysis of social data basedon emotion-bearing patterns.SocialNLP 2014, page 38, 2014.
[4] A. Betella and P. F. Verschure. The affective slider: A digital self-assessment scalefor the measurement of human emotions.PloS one, 11(2):e0148037, 2016.
[5] C. M. Bishop. Pattern recognition.Machine Learning, 128:110–113, 2006.
[6] J. Boyd-Graber and P. Resnik. Holistic sentiment analysis across languages: Mul-tilingual supervised latent dirichlet allocation. InProceedings of the 2010 Confer-ence on Empirical Methods in Natural Language Processing, pages 45–55. Asso-ciation for Computational Linguistics, 2010.
[7] M. M. Bradley and P. J. Lang. Measuring emotion: the self-assessment manikinand the semantic differential.Journal of behavior therapy and experimental psy-chiatry, 25(1):49–59, 1994.
[8] M. M. Bradley and P. J. Lang. Affective norms for english words (anew): In-struction manual and affective ratings. Technical report, Technical report C-1, thecenter for research in psychophysiology, University of Florida, 1999.
[9] M. Buhrmester, T. Kwang, and S. D. Gosling. Amazon’s mechanical turk a newsource of inexpensive, yet high-quality, data?Perspectives on psychological sci-ence, 6(1):3–5, 2011.
[10] C. Callison-Burch. Fast, cheap, and creative: evaluating translation quality usingamazon’s mechanical turk. InProceedings of the 2009 Conference on EmpiricalMethods in Natural Language Processing: Volume 1-Volume 1, pages 286–295.Association for Computational Linguistics, 2009.
[11] R. A. Calvo and S. D’Mello. Affect detection: An interdisciplinary review of mod-els, methods, and their applications.IEEE Transactions on affective computing,1(1):18–37, 2010.
[12] M. J. Crump, J. V. McDonnell, and T. M. Gureckis. Evaluating amazon’s mechan-ical turk as a tool for experimental behavioral research.PloS one, 8(3):e57410,2013.
[13] P. Ekman. An argument for basic emotions.Cognition & emotion, 6(3-4):169–200, 1992.
[14] P. Ekman, E. R. Sorenson, W. V. Friesen, et al. Pan-cultural elements in facialdisplays of emotion.Science, 164(3875):86–88, 1969.
[15] A.-L. Gilet and C. Jallais. Valence, arousal and word associations.Cognition andEmotion, 25(4):740–746, 2011.
[16] J. R. Hershey and P. A. Olsen. Approximating the kullback leibler divergencebetween gaussian mixture models. InAcoustics, Speech and Signal Processing,2007. ICASSP 2007. IEEE International Conference on, volume 4, pages IV–317.IEEE, 2007.
[17] R. James A. A circumplex model of affect.Journal of Personality and SocialPsychology, pages 1161–1178, 1980.
[18] O. P. John and S. Srivastava. The big five trait taxonomy: History, measurement,and theoretical perspectives.Handbook of personality: Theory and research,2(1999):102–138, 1999.
[19] S. Kim, J. Bak, and A. H. Oh. Do you feel what i feel? social aspects of emotionsin twitter conversations. InICWSM, 2012.
[20] Y. Kim. Convolutional neural networks for sentence classification.arXiv preprintarXiv:1408.5882, 2014.
[21] S. Korkmaz, D. Goksuluk, and G. Zararsiz. Mvn: an r package for assessingmultivariate normality.The R Journal, 6(2):151–162, 2014.
[22] E. Kouloumpis, T. Wilson, and J. D. Moore. Twitter sentiment analysis: The goodthe bad and the omg!Icwsm, 11:538–541, 2011.
[23] P. Kuppens, F. Tuerlinckx, M. Yik, P. Koval, J. Coosemans, K. J. Zeng, and J. A.Russell. The relation between valence and arousal in subjective experience varieswith personality and culture.Journal of personality, 2016.
[24] A. L. Maas, R. E. Daly, P. T. Pham, D. Huang, A. Y. Ng, and C. Potts. Learningword vectors for sentiment analysis. InProceedings of the 49th Annual Meeting ofthe Association for Computational Linguistics: Human Language Technologies-Volume 1, pages 142–150. Association for Computational Linguistics, 2011.
[25] K. V. Mardia. Measures of multivariate skewness and kurtosis with applications.Biometrika, pages 519–530, 1970.
[26] G. Matthews, I. J. Deary, and M. C. Whiteman.Personality traits. CambridgeUniversity Press, 2003.
[27] T. Mikolov, K. Chen, G. Corrado, and J. Dean. Efficient estimation of word rep-resentations in vector space.arXiv preprint arXiv:1301.3781, 2013.
[28] T. Mikolov, I. Sutskever, K. Chen, G. S. Corrado, and J. Dean. Distributed repre-sentations of words and phrases and their compositionality. InAdvances in neuralinformation processing systems, pages 3111–3119, 2013.
[29] F. ̊A. Nielsen. A new anew: Evaluation of a word list for sentiment analysis inmicroblogs.arXiv preprint arXiv:1103.2903, 2011.
[30] F. Nijboer, F. O. Morin, S. P. Carmien, R. A. Koene, E. Leon, and U. Hoffmann.Affective brain-computer interfaces: Psychophysiological markers of emotion inhealthy persons and in persons with amyotrophic lateral sclerosis. InAffectiveComputing and Intelligent Interaction and Workshops, 2009. ACII 2009. 3rd In-ternational Conference on, pages 1–11. IEEE, 2009.
[31] C. E. Osgood. The nature and measurement of meaning.Psychological bulletin,49(3):197, 1952.
[32] A. Pak and P. Paroubek. Twitter as a corpus for sentiment analysis and opinionmining. InLREc, volume 10, pages 1320–1326, 2010.
[33] G. Paltoglou and M. Thelwall. Seeing stars of valence and arousal in blog posts.IEEE Transactions on Affective Computing, 4(1):116–123, 2013.
[34] B. Pang, L. Lee, et al. Opinion mining and sentiment analysis.Foundations andTrendsR©in Information Retrieval, 2(1–2):1–135, 2008.
[35] R. 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.
[36] J. Posner, J. A. Russell, and B. S. Peterson. The circumplex model of affect: Anintegrative approach to affective neuroscience, cognitive development, and psy-chopathology.Development and psychopathology, 17(03):715–734, 2005.
[37] D. Preotiuc-Pietro, H. A. Schwartz, G. Park, J. C. Eichstaedt, M. Kern, L. Un-gar, and E. P. Shulman. Modelling valence and arousal in facebook posts. InProceedings of NAACL-HLT, pages 9–15, 2016.
[38] K. R. Scherer. What are emotions? and how can they be measured?Social scienceinformation, 44(4):695–729, 2005.
[39] K. Sun, J. Yu, Y. Huang, and X. Hu. An improved valence-arousal emotionspace for video affective content representation and recognition. InMultimediaand Expo, 2009. ICME 2009. IEEE International Conference on, pages 566–569.IEEE, 2009.
[40] S. Volkova, T. Wilson, and D. Yarowsky. Exploring sentiment in social media:Bootstrapping subjectivity clues from multilingual twitter streams. InACL (2),pages 505–510, 2013.
[41] X. Wan. Co-training for cross-lingual sentiment classification. InProceedings ofthe Joint Conference of the 47th Annual Meeting of the ACL and the 4th Interna-tional Joint Conference on Natural Language Processing of the AFNLP: Volume1-volume 1, pages 235–243. Association for Computational Linguistics, 2009.
[42] H. Wang, D. Can, A. Kazemzadeh, F. Bar, and S. Narayanan. A system for real-time twitter sentiment analysis of 2012 us presidential election cycle. InProceed-ings of the ACL 2012 System Demonstrations, pages 115–120. Association forComputational Linguistics, 2012.
[43] J. Wang, L.-C. Yu, K. R. Lai, and X. Zhang. Community-based weighted graphmodel for valence-arousal prediction of affective words.IEEE/ACM Transactionson Audio, Speech, and Language Processing, 24(11):1957–1968, 2016.
[44] J.-C. Wang, Y.-H. Yang, H.-M. Wang, and S.-K. Jeng. Modeling the affectivecontent of music with a gaussian mixture model.IEEE Transactions on AffectiveComputing, 6(1):56–68, 2015.
[45] A. B. Warriner, V. Kuperman, and M. Brysbaert. Norms of valence, arousal, anddominance for 13,915 english lemmas.Behavior research methods, 45(4):1191–1207, 2013.
[46] D. Watson, W. Gamez, and L. J. Simms. Basic dimensions of temperament andtheir relation to anxiety and depression: A symptom-based perspective.Journalof Research in Personality, 39(1):46–66, 2005.
[47] Y.-H. Yang and J.-Y. Liu. Quantitative study of music listening behavior in a socialand affective context.Multimedia, IEEE Transactions on, 15(6):1304–1315, 2013.
[48] S. Zhao, H. Yao, and X. Jiang. Predicting continuous probability distribution ofimage emotions in valence-arousal space. InProceedings of the 23rd Annual ACMConference on Multimedia Conference, pages 879–882. ACM, 2015.