臺灣博碩士論文加值系統

現代人因為科技日趨發達，遇到挫折、壓力而情緒較為負面時，習慣透過瀏覽器到網路上搜尋與自己問題相關的文章來當作參考，但傳統瀏覽器只能以關鍵字做搜尋，且網路上文章太多太雜，要找到符合自己問題又有參考價值之文章常需要耗費大量時間。為了解決上述問題，本論文提出整合機器學習與情感運算於聊天機器人架構(Integrating Machine Learning and Affective Computing into Chatbot Framework，下方簡稱IMACF)。本論文以社群網站Dcard上的文章作為訓練資料，共會建立三種模型，分別為用於情感運算的二元分類(Binary Classification)模型、分析話題的多元分類(Multi-class Classification)模型及推薦文章的潛在語義索引(Latent Semantic Indexing)模型，並基於IMACF建立情感運算與文章推薦平台(Affective Computing and Article Recommendation Platform，下方簡稱ACARP)來讓使用者使用，以此驗證IMACF之可行性。首先本論文使用TF-IDF引導使用者與聊天機器人進行對話來描述自己的問題，並透過機器學習自動判斷對話文本是否已經足夠讓系統進行後續分析，當對話文本足夠後便會將文本進行斷詞、詞嵌入等步驟轉換為可以進行機器學習的數字向量，接著將此向量輸入到二元分類模型中進行情感運算，並分析出使用者之情緒(Emotion)狀態;而為了能夠準確的推薦與使用者相關的文章，本論文以多元分類及Ontology UML概念來分析使用者之話題領域，最後以分析結果為基準，使用潛在語義索引模型來推薦相關的文章供其參考，以此大幅提升使用者的搜尋效率，並改善只能以關鍵字搜尋的缺點。本論文系統具有可擴充性，因此未來會加入更多不同話題的文章來當作訓練資料，為了提升效率，本論文使用Spark叢集架構來加快模型訓練的速度，並參考邊緣運算(Edge Computing)的概念來將訓練模型、載入模型的架構分開運算，以此提高整體系統的執行效率。本論文為了評估ACARP在真實情況下是否能夠符合使用者的需求，做了系統實際使用回饋統計，並根據使用者的回饋內容來調整系統模型的參數，藉此訓練出真正讓使用者滿意的模型。

Modern people often rely on Internet to search for any reference articles related to their own problems of relative negatively emotions. However, traditional browsers will only search with keywords that it is time-consuming for users to identify ones with reference value among a large variety of articles. In order to solve the above problem, this paper proposed the Integrated Machine Learning and Affective Computing into Chatbot Framework (IMACF), which integrates machine learning and affective computing into chatbot. In this paper, we use the articles on the community website Dcard as training materials to establish a machine learning model and an Affective Computing and Article Recommendation Platform (ACARP) based on IMACF for users, as a way to verify the feasibility of IMACF. First, this paper adopted TF-IDF to guide users to have a dialogue with chatbot to describe their own problems. When the dialogue texts provided sufficient information, the texts would be converted into a digital vector for machine learning by steps such as Word Segmentation and Word Embedding, and input into a binary classification model for emotional operation to analyze the emotional state of users. In order to accurately recommend articles related to users, this paper used multivariate classification and Ontology UML concept to analyze users' discussion topics, where the latent semantic index model could then recommend related articles for their reference. The system in this paper was scalable, as more articles on different topics could be added as training data in the future. To improve efficiency, Spark cluster architecture and the concept of Edge Computing would help to improve the execution efficiency of the overall system. In order to evaluate whether ACARP can meet the needs of users in real situations, this study relied on the feedback statistics of the actual system use, and adjusted the parameters of the system model according to the feedback content of users, to train a model to truly reflect and satisfy users.

摘要...........................i
Abstract...........................ii
誌謝...........................iii
目錄...........................iv
表目錄...........................vii
圖目錄...........................viii
第一章 簡介...........................1
1.1 研究背景...........................1
1.2 研究動機...........................2
1.3 研究目的...........................2
第二章 研究技術探討...........................3
2.1 聊天機器人...........................3
2.2 CKIP Tagger斷詞系統...........................4
2.3 Jieba斷詞系統...........................5
2.4 Python Flask...........................5
2.5 潛在語義索引(Latent Semantic Indexing).........6
2.6 TF-IDF...........................7
2.7 二元分類...........................8
2.7.1 決策樹(Decision Tree).........................8
2.7.2 邏輯迴歸(Logistic Regression)....................9
2.7.3 支援向量機器(Support Vector Machine)............10
2.7.4 梯度提升樹(Gradient Boosting Decision Tree).....11
2.8 多元分類...........................11
2.9 混淆矩陣與評估指標...........................12
2.10 Hadoop...........................13
2.10.1 Hadoop分散式檔案系統...........................14
2.10.2 Hadoop YARN...........................15
2.11 Spark...........................16
2.11.1 Spark彈性分散式資料集...........................17
2.11.2 Spark叢集架構及管理器...........................18
2.12 邊緣運算(Edge Computing)...........................19
2.13 相關研究...........................20
第三章 系統設計與開發步驟...........................21
3.1 系統設計...........................21
3.1.1 系統模組說明...........................21
3.1.2 系統運作流程說明...........................23
3.1.3 系統開發環境...........................24
3.2 研究方法與流程...........................25
3.2.1 資料蒐集模組...........................26
3.2.2 資料前處理模組...........................27
3.2.3 情感運算模組...........................29
3.2.4 建立LSI模型...........................32
3.2.5 TF-IDF整合LSI引導對話...........................35
3.2.6 推薦文章模組...........................39
3.2.7 系統架構模組化...........................43
3.2.8 聊天機器人整合模組...........................44
3.2.9 雲端運算與邊緣運算...........................44
第四章 系統應用場景與展示...........................45
4.1 應用場景...........................45
4.2 系統特色...........................45
4.3 系統運行實際狀況...........................46
第五章 系統實驗與準確度評估...........................50
5.1 雲端運算環境架構...........................50
5.1.1 測試環境...........................50
5.1.2 測試環境架構...........................52
5.2 效能評估...........................53
5.3 準確性評估...........................55
5.3.1 機器學習演算法準確度評估...........................55
5.3.2 深度學習演算法準確度評估...........................58
5.4 實驗結論與系統評估...........................60
5.4.1 情感運算準確度測試...........................60
5.4.2 終止對話門檻值測試...........................61
5.4.3 文章推薦K值測試...........................62
5.4.4 真實測驗結論...........................65
第六章 結論...........................66
參考文獻...........................67
Extended Abstract...........................71
I............................INTRODUCTION 72
II. BACKROUND...........................72
A. Chatbot...........................72
B. Latent Semantic Indexing...........................72
C. CKIP Tagger...........................72
D. TF-IDF...........................73
E. Binary Classification...........................73
F. Spark...........................73
G. Edge Computing...........................73
H. Related Works...........................73
III. SYSTEM DESIGN AND PROCEDURES..................74
A. System Design...........................74
B. System Architecture...........................75
IV. SYSTEM DEVELOPMENT...........................76
A. Data Collection Module...........................76
B. Data Preparation Module...........................76
C. Affective Computing Module...........................77
D. Latent Semantic Indexing Module...................77
E. Data Analysis...........................79
F. Guidance Dialogue...........................79
G. System Demonstration...........................80
V. SYSTEM DESIGN AND PROCEDURES.......................81
A. Spark Cluster Environment...........................81
B. Performance Test...........................82
C. F1-Score Test...........................82
D. Actual Case Test and Result...........................83
VI. CONCLUSION...........................85
REFERENCE...........................86

[1]Wikipedia. Emotion. Available: https://en.wikipedia.org/wiki/Emotion
[2]A. Banafa. What is Affective Computing? Available: https://www.bbvaopenmind.com/en/technology/digital-world/what-is-affective-computing/
[3]R. Boorugu and G. Ramesh, "A Survey on NLP based Text Summarization for Summarizing Product Reviews," in 2020 Second International Conference on Inventive Research in Computing Applications (ICIRCA), 2020, pp. 352-356: IEEE.
[4]i. m. hub. (2020). Best AI Chatbot Platforms for 2021. Available: https://influencermarketinghub.com/ai-chatbot-platforms/
[5]Wikipedia. (2020). Natural-language understanding. Available: https://en.wikipedia.org/wiki/Natural-language_understanding
[6]ITStar. 何必糾結hadoop和spark如何選擇. Available: https://kknews.cc/zh-tw/tech/e8m2eoz.html
[7]P. Miller. What is edge computing? Available: https://www.theverge.com/circuitbreaker/2018/5/7/17327584/edge-computing-cloud-google-microsoft-apple-amazon
[8]Wikipedia. Chatbot. Available: https://en.wikipedia.org/wiki/Chatbot
[9]S. Singh and H. K. Thakur, "Survey of Various AI Chatbots Based on Technology Used," in 2020 8th International Conference on Reliability, Infocom Technologies and Optimization (Trends and Future Directions)(ICRITO), 2020, pp. 1074-1079: IEEE.
[10]Google. Dialogflow Documentation. Available: https://cloud.google.com/dialogflow/docs
[11]O. Keszocze and I. G. Harris, "Chatbot-based assertion generation from natural language specifications," in 2019 Forum for Specification and Design Languages (FDL), 2019, pp. 1-6: IEEE.
[12]P.-H. Li, T.-J. Fu, and W.-Y. Ma, "Why Attention? Analyze BiLSTM Deficiency and Its Remedies in the Case of NER," in AAAI, 2020, pp. 8236-8244.
[13]CKIP. CkipTagger. Available: https://github.com/ckiplab/ckiptagger
[14]APCLab. Jieba-tw. Available: https://github.com/APCLab/jieba-tw
[15]P.-J. Lin, S. Hung, S. F. S. Lam, and B. C. Chen, "Object Recognition with Machine Learning: Case Study of Demand-Responsive Service," in 2019 IEEE International Conference on Internet of Things and Intelligence System (IoTaIS), 2019, pp. 129-134: IEEE.
[16]Wikipedia. (2020). 潛在語義索引. Available: https://zh.wikipedia.org/wiki/%E6%BD%9C%E5%9C%A8%E8%AF%AD%E4%B9%89%E7%B4%A2%E5%BC%95
[17]M. J. L. a. Brand and i. applications, "Fast low-rank modifications of the thin singular value decomposition," vol. 415, no. 1, pp. 20-30, 2006.
[18]Wikipedia. tf-idf. Available: https://en.wikipedia.org/wiki/Tf%E2%80%93idf
[19]J. Brownlee. (2020). 4 Types of Classification Tasks in Machine Learning. Available: https://machinelearningmastery.com/types-of-classification-in-machine-learning/
[20]N. Rochmawati et al., "Covid Symptom Severity Using Decision Tree," in 2020 Third International Conference on Vocational Education and Electrical Engineering (ICVEE), 2020, pp. 1-5: IEEE.
[21]P. Mathur, S. K. Khatri, and M. Sharma, "Prediction of aviation accidents using logistic regression model," in 2017 International Conference on Infocom Technologies and Unmanned Systems (Trends and Future Directions)(ICTUS), 2017, pp. 725-728: IEEE.
[22]Wikipedia. (2021). Dependent and independent variables. Available: https://en.wikipedia.org/wiki/Dependent_and_independent_variables
[23]L. Tyagi, V. Kumar, and S. Chakraborty, "Explosion Consequence Analysis for Military Targets Through Support Vector Machines," in 2020 8th International Conference on Reliability, Infocom Technologies and Optimization (Trends and Future Directions)(ICRITO), 2020, pp. 948-951: IEEE.
[24]R. Gandhi. (2018). Support Vector Machine — Introduction to Machine Learning Algorithms. Available: https://towardsdatascience.com/support-vector-machine-introduction-to-machine-learning-algorithms-934a444fca47
[25]S. Manna, S. Biswas, R. Kundu, S. Rakshit, P. Gupta, and S. Barman, "A statistical approach to predict flight delay using gradient boosted decision tree," in 2017 International Conference on Computational Intelligence in Data Science (ICCIDS), 2017, pp. 1-5: IEEE.
[26]C. Maklin. (2019). Gradient Boosting Decision Tree Algorithm Explained. Available: https://towardsdatascience.com/machine-learning-part-18-boosting-algorithms-gradient-boosting-in-python-ef5ae6965be4
[27]G. Campbell, M. Douglas, and J. Bailey, "Nonparametric comparison of two tests of cardiac function on the same patient population using the entire ROC curve," in Proceedings. Computers in Cardiology 1988, 1988, pp. 267-270: IEEE.
[28]S. Narkhede. (2018). Understanding Confusion Matrix. Available: https://towardsdatascience.com/understanding-confusion-matrix-a9ad42dcfd62
[29]M. Bhandarkar, "MapReduce programming with apache Hadoop," in 2010 IEEE International Symposium on Parallel & Distributed Processing (IPDPS), 2010, pp. 1-1: IEEE Computer Society.
[30]G. Yang, "The application of MapReduce in the cloud computing," in 2011 2nd International Symposium on Intelligence Information Processing and Trusted Computing, 2011, pp. 154-156: IEEE.
[31]Y. Kim, T. Araragi, J. Nakamura, and T. Masuzawa, "A distributed NameNode cluster for a highly-available Hadoop distributed file system," in 2014 IEEE 33rd International Symposium on Reliable Distributed Systems, 2014, pp. 333-334: IEEE.
[32]T. H. Aung and W. T. Zaw, "Improved Job Scheduling for Achieving Fairness on Apache Hadoop YARN," in 2020 International Conference on Advanced Information Technologies (ICAIT), 2020, pp. 188-193: IEEE.
[33]G. Gousios, "Big data software analytics with Apache Spark," in Proceedings of the 40th International Conference on Software Engineering: Companion Proceeedings, 2018, pp. 542-543.
[34]K. Aziz, D. Zaidouni, and M. Bellafkih, "Real-time data analysis using Spark and Hadoop," in 2018 4th International Conference on Optimization and Applications (ICOA), 2018, pp. 1-6: IEEE.
[35]Y.-C. Hsieh et al., "Managed edge computing on Internet-of-Things devices for smart city applications," in NOMS 2018-2018 IEEE/IFIP Network Operations and Management Symposium, 2018, pp. 1-2: IEEE.
[36]P. B. Kulkarni and M. M. Patil, "Clinical Depression Detection in Adolescent by Face," in 2018 International Conference on Smart City and Emerging Technology (ICSCET), 2018, pp. 1-4: IEEE.
[37]R. W. Picard, Affective computing. MIT press, 2000.
[38]E. J. I. i. s. Cambria, "Affective computing and sentiment analysis," vol. 31, no. 2, pp. 102-107, 2016.
[39]A. Yaganteeswarudu, "Multi Disease Prediction Model by using Machine Learning and Flask API," in 2020 5th International Conference on Communication and Electronics Systems (ICCES), 2020, pp. 1242-1246: IEEE.
[40]M. Li, Q. Lu, Y. Long, and L. J. I. T. o. A. C. Gui, "Inferring affective meanings of words from word embedding," vol. 8, no. 4, pp. 443-456, 2017.
[41]F. Patel, R. Thakore, I. Nandwani, and S. K. Bharti, "Combating Depression in Students using an Intelligent ChatBot: A Cognitive Behavioral Therapy," in 2019 IEEE 16th India Council International Conference (INDICON), 2019, pp. 1-4: IEEE.
[42]M. R. Hasan, M. Maliha, and M. Arifuzzaman, "Sentiment Analysis with NLP on Twitter Data," in 2019 International Conference on Computer, Communication, Chemical, Materials and Electronic Engineering (IC4ME2), 2019, pp. 1-4: IEEE.
[43]A. V. Hazarika, G. J. S. R. Ram, and E. Jain, "Performance comparision of Hadoop and spark engine," in 2017 International Conference on I-SMAC (IoT in Social, Mobile, Analytics and Cloud)(I-SMAC), 2017, pp. 671-674: IEEE.
[44]C.-F. Wang. (2019). The Vanishing Gradient Problem. Available: https://towardsdatascience.com/the-vanishing-gradient-problem-69bf08b15484
[45]Y. huang. (2017). 以 jieba 探索文本主題 (歌詞文本). Available: https://github.com/youngmihuang/lyrics_application/blob/master/lyrics_2.ipynb
[46]林大貴, TensorFlow+ Keras 深度學習人工智慧實務應用. 博碩文化股份有限公司, 2017.