在現代生活中，很多人正在遭受心理疾病的困擾。隨著互聯網服務的發展，出現了一些專業的線上醫療服務平臺。選擇通過網路管道傾訴心理問題，以獲取專業人員提供的解決辦法，已經成為一種既方便又顧忌用戶隱私的有效途徑。互聯網醫療服務平臺積累了的大量社會心理學文本，如果能有效挖掘這一寶貴的資源，將會給學術研究來新的突破方向，通過運用這些研究的成果，互聯網服務平臺也可以給用户者提供更好的體驗。
心理疾病文本屬於醫學文本，通常對一個病情的描述會被歸納為多個情緒標籤，屬於多標籤文本。識別文本情緒標籤是自然語言處理中情感分析任務的一個研究方向，本文探討心理疾病領域的多標籤文本情緒識別問題，再此基礎上重點探討了對修飾結構的對多標籤分類的影響。在情感分析研究中，對情感詞彙的研究一直是重點，其是決定文本情緒的分類的重要特徵。本文在關注文本中情緒詞的同時結合修飾詞的分析，構建了修飾結構詞典，通過詞典對文本進行特徵提取，分析了含有修飾結構特徵的文本對多標籤情緒識別的影響。本文考慮了三種類別的修飾詞，否定詞、程度副詞和虛詞，由這些類別的修飾詞與情感片語合，構成不同的修飾結構。否定修飾結構，可以用於扭轉結構中情感詞的語義極性；程度修飾結構則可以增強或者減弱情緒詞的正或負程度；虛詞修飾結構可以減弱情緒強度。
本文首先探討了使用深度學習的模型進行自然語言領域文本分類任務的可行性，提出了基於句子層面的幾種深度神經網路的分類模型。其次，使用分類模型進行心理疾病文本多標籤分類實驗。接著，搭建修飾結構詞典，用該詞典提取含有修飾結構的特徵文。最後，對特徵文本使用分類模型進行分類預測，評估修飾結構對心理疾病文本情緒識別準確性的影響。

In modern life, many people are suffering from mental illness problems. With the development of Internet services, there have been some professional online medical service platform. Choose to talk through the network through the psychological problems, in order to obtain the solution provided by professionals, has become a convenient and scruples the user an effective way of privacy. Internet medical service platform has accumulated a large number of social psychology text, if you can effectively tap this valuable resource, will give academic research to a new breakthrough direction, through the use of the results of these studies, the Internet service platform can also provide users with A better experience.
Mental disease text is a medical text, usually a description of a condition will be summarized as multiple emotional tags, are multi-label text. It is a research direction of emotional analysis task in natural language processing. This paper discusses the problem of multi-label text emotion recognition in the field of mental illness, and then discusses the influence of modified structure on multi-label classification. In the study of emotional analysis, the study of emotional vocabulary has always been the focus, which is the important feature of the classification of text emotions. This paper analyzes the influence of the text containing the modified structure on the multi - label emotion recognition by analyzing the feature of the text by using the dictionary to construct the modified structure dictionary. This article considers three kinds of modifiers, negative words, degree adverbs and functional words, which are composed of different types of modifiers and emotional films. Negative modification structure can be used to reverse the semantic polarity of emotion words in the structure. The degree of modification structure can enhance or weaken the positive or negative degree of emotional words. The function structure can weaken the emotional intensity.
This paper first discusses the feasibility of using the depth learning model to classify text in the natural language field, and proposes a classification model of several depth neural networks based on sentence level. Secondly, the classification model was used to carry out the multi - label classification experiment of mental illness text. Then, the structure of the modified structure dictionary, with the dictionary to extract the text containing the modified structure. Finally, the classification text of the feature text is classified and predicted, and the influence of the modified structure on the accuracy of the text emotion recognition is evaluated.

目錄
摘要 iii
ABSTRACT iv
致謝 vi
目錄 vii
表目錄 ix
圖目錄 x
第一章 引言 1
1.1研究動機 1
1.2研究挑戰 2
1.3研究貢獻 3
1.4組織結構 3
第二章 背景和相關工作 4
2.1文本情感分析 4
2.1.1情感分析目標 4
2.1.2情感分析層次 4
2.2修飾結構 5
2.3多標籤分類 6
2.4詞嵌入技術 7
2.4.1詞的向量化表示 7
2.4.2向量訓練工具 8
2.5深度神經網路模型用於文本分類 9
2.5.1多層感知器 9
2.5.2卷積神經網路 10
2.5.3循環神經網絡 12
2.5.4長短期記憶神經網路 13
第三章 實驗方法 15
3.1文本多標籤分類 15
3.2文本修飾結構特徵提取 16
3.3用於文本分類的神經網路模型 17
3.3.1 卷積神經網路模型(CNN) 19
3.3.2 多通道卷積神經網路（Multichannel-CNN） 19
3.3.3 雙向長短期記憶和卷積神經網路的組合模型（Bi-LSTM_CNN） 20
第四章 實驗及結果分析 22
4.1實驗資料 22
4.2實驗步驟 23
4.2.1實驗資料處理 23
4.2.2訓練詞向量 25
4.2.3搭建訓練模型 25
4.2.4訓練模型 26
4.3評估方法 26
4.4實驗設定 27
4.5實驗結果及分析 28
第五章 總結 34
參考資料 35

[1] Liu, Bing. "Sentiment analysis and opinion mining." Synthesis lectures on human language technologies 5.1 (2012): 1-167.
[2] Liu, Tie-Yan. "Learning to rank for information retrieval." Foundations and Trends® in Information Retrieval 3.3 (2009): 225-331.
[3] Brostedt, E. M., and N. L. Pedersen. "Stressful life events and affective illness." Acta Psychiatrica Scandinavica 107.3 (2003): 208-215.
[4] Miller, George, and Christiane Fellbaum. "Wordnet: An electronic lexical database." (1998).
[5] Yu, Liang-Chih, and Chun-Yuan Ho. "Identifying Emotion Labels from Psychiatric Social Texts Using Independent Component Analysis." COLING. 2014.
[6] Pink, Glen, Will Radford, and Ben Hachey. "Classification of mental health forum posts." CLPsych@ HLT-NAACL. 2016.
[7] 儲澤祥, and 謝曉明. "漢語語法化研究中應重視的若干問題." 世界漢語教學 2 (2002): 5-13.
[8] Pang, Bo, and Lillian Lee. "Opinion mining and sentiment analysis." Foundations and Trends® in Information Retrieval 2.1–2 (2008): 1-135.
[9] Calvo, Rafael A., and Sidney D'Mello. "Affect detection: An interdisciplinary review of models, methods, and their applications." IEEE Transactions on affective computing 1.1 (2010): 18-37.
[10] Liu, Bing. "Sentiment analysis and opinion mining." Synthesis lectures on human language technologies 5.1 (2012): 1-167.
[11] Feldman, Ronen. "Techniques and applications for sentiment analysis." Communications of the ACM 56.4 (2013): 82-89.
[12] Ekman, Paul. "An argument for basic emotions." Cognition & emotion 6.3-4 (1992): 169-200.
[13] Rosenthal, Sara, et al. "SemEval-2015 Task 10: Sentiment Analysis in Twitter." SemEval@ NAACL-HLT. 2015.
[14] Russell, James A. "Core affect and the psychological construction of emotion." Psychological review 110.1 (2003): 145.
[15] Pennebaker, J. W., R. J. Booth, and M. E. Francis. "Linguistic Inquiry and Word Count: LIWC [computer program]." Austin, TX: LIWC. net (2007).
[16] Bradley, Margaret M., and Peter J. Lang. Affective norms for English words (ANEW): Instruction manual and affective ratings. Technical report C-1, the center for research in psychophysiology, University of Florida, 1999.
[17] Huang, Chin-Lan, et al. "The development of the Chinese linguistic inquiry and word count dictionary." Chinese Journal of Psychology 54.2 (2012): 185-201.
[18] Yu, Liang-Chih, et al. "Building Chinese Affective Resources in Valence-Arousal Dimensions." HLT-NAACL. 2016.
[19] Kiritchenko, Svetlana, and Saif M. Mohammad. "Capturing Reliable Fine-Grained Sentiment Associations by Crowdsourcing and Best-Worst Scaling." HLT-NAACL. 2016.
[20] Kiritchenko, Svetlana, Saif Mohammad, and Mohammad Salameh. "SemEval-2016 Task 7: Determining Sentiment Intensity of English and Arabic Phrases." SemEval@ NAACL-HLT. 2016.
[21] Kennedy, Alistair, and Diana Inkpen. "Sentiment classification of movie reviews using contextual valence shifters." Computational intelligence 22.2 (2006): 110-125.
[22] Clare, Amanda, and Ross King. "Knowledge discovery in multi-label phenotype data." Principles of data mining and knowledge discovery (2001): 42-53.
[23] Zhang, Min-Ling, and Zhi-Hua Zhou. "A review on multi-label learning algorithms." IEEE transactions on knowledge and data engineering 26.8 (2014): 1819-1837.
[24] Luaces, Oscar, et al. "Binary relevance efficacy for multilabel classification." Progress in Artificial Intelligence (2012): 1-11.
[25] Zhang, Min-Ling, and Zhi-Hua Zhou. "ML-KNN: A lazy learning approach to multi-label learning." Pattern recognition 40.7 (2007): 2038-2048.
[26] Tsoumakas, Grigorios, Ioannis Katakis, and Ioannis Vlahavas. "Mining multi-label data." Data mining and knowledge discovery handbook (2010): 667-685.
[27] Hinton, Geoffrey E. "Learning distributed representations of concepts." Proceedings of the eighth annual conference of the cognitive science society. Vol. 1. 1986.
[28] Lan, Man, et al. "Three Convolutional Neural Network-based models for learning Sentiment Word Vectors towards sentiment analysis." Neural Networks (IJCNN), 2016 International Joint Conference on. IEEE, 2016.
[29] Bengio, Yoshua, et al. "A neural probabilistic language model." Journal of machine learning research 3.Feb (2003): 1137-1155.
[30] Mikolov, Tomas, et al. "Efficient estimation of word representations in vector space." arXiv preprint arXiv:1301.3781 (2013).
[31] Mikolov, Tomas, et al. "Distributed representations of words and phrases and their compositionality." Advances in neural information processing systems. 2013.
[32] Pennington, Jeffrey, Richard Socher, and Christopher D. Manning. "Glove: Global vectors for word representation." EMNLP. Vol. 14. 2014.
[33] Go, Alec, Richa Bhayani, and Lei Huang. "Twitter sentiment classification using distant supervision." CS224N Project Report, Stanford 1.2009 (2009): 12.
[34] Conway, Mike, et al. "Classifying disease outbreak reports using n-grams and semantic features." International journal of medical informatics 78.12 (2009): e47-e58.
[35] Tan, Chade-Meng, Yuan-Fang Wang, and Chan-Do Lee. "The use of bigrams to enhance text categorization." Information processing & management 38.4 (2002): 529-546.
[36] Lertnattee, Verayuth, and Thanaruk Theeramunkong. "Multidimensional text classification for drug information." IEEE Transactions on Information Technology in Biomedicine 8.3 (2004): 306-312.
[37] Kononenko, Igor. "Semi-naive Bayesian classifier." Machine Learning—EWSL-91. Springer Berlin/Heidelberg, 1991.
[38] Kiros, Ryan, et al. "Skip-thought vectors." Advances in neural information processing systems. 2015.
[39] Iyyer, Mohit, et al. "Deep Unordered Composition Rivals Syntactic Methods for Text Classification." ACL (1). 2015.
[40] Kim, Yoon. "Convolutional neural networks for sentence classification." arXiv preprint arXiv:1408.5882 (2014).
[41] Kalchbrenner, Nal, Edward Grefenstette, and Phil Blunsom. "A convolutional neural network for modelling sentences." arXiv preprint arXiv:1404.2188(2014).
[42] Graves, Alex. Supervised sequence labelling with recurrent neural networks. Vol. 385. Springer Science & Business Media, 2012.
[43] Irsoy, Ozan, and Claire Cardie. "Opinion Mining with Deep Recurrent Neural Networks." EMNLP. 2014.
[44] Dyer, Chris, et al. "Transition-based dependency parsing with stack long short-term memory." arXiv preprint arXiv:1505.08075 (2015).
[45] Dos Santos, Cícero Nogueira, and Maira Gatti. "Deep Convolutional Neural Networks for Sentiment Analysis of Short Texts." COLING. 2014.
[46] Yin, Wenpeng, and Hinrich Schütze. "Multichannel variable-size convolution for sentence classification." arXiv preprint arXiv:1603.04513 (2016).
[47] Liu, Pengfei, Shafiq R. Joty, and Helen M. Meng. "Fine-grained Opinion Mining with Recurrent Neural Networks and Word Embeddings." EMNLP. 2015.
[48] Hochreiter, Sepp, and Jürgen Schmidhuber. "Long short-term memory." Neural computation 9.8 (1997): 1735-1780.
[49] Kiritchenko, Svetlana, and Saif Mohammad. "The Effect of Negators, Modals, and Degree Adverbs on Sentiment Composition." WASSA@ NAACL-HLT. 2016.
[50] Cavnar, William B., and John M. Trenkle. "N-gram-based text categorization." Ann Arbor MI 48113.2 (1994): 161-175.
[51] Saif, Hassan, et al. "On stopwords, filtering and data sparsity for sentiment analysis of twitter." (2014): 810-817.
[52] Dendamrongvit, Sareewan, and Miroslav Kubat. "Undersampling Approach for Imbalanced Training Sets and Induction from Multi-label Text-Categorization Domains." PAKDD Workshops. 2009.