聊天機器人是一種能夠提供各項服務的即時訊息代理人，在近來已成為一種日益流行的話題，並廣泛應用於各個領域。而隨著健康意識的提高，人們開始關注健康問題，並且也越來越多聊天機器人在醫療保健中的應用。雖然在衛教中已經有許多聊天機器人應用案例，但大多數都還是使用衛教文章或影片作為回應，他們無法簡單而直接精準的回答問題。這使得使用者無法得到立即答案。
因此，為了解決上述問題，本研究提出了一種養生保健建議提取模型，從線上醫療諮詢平台的數據集中提取養生保健建議，並構建疾病知識圖譜。基於知識圖譜，我們提出了一個養生保健建議聊天機器人，可以理解自然語言並為用戶提供養生保健建議。
實驗中證明透過語料庫的成長幅度會隨著資料量增加而趨於收斂。而模型的正確率在測試資料達到95.7%。

A chatbot is an instant messaging agent that provides services, which has become an increasingly popular topic and is widely used in various fields. As rising health awareness, people are paying attention to their health, and more chatbots in healthcare are being developed. Although there are already many chatbot applications in patient education, most of them are using patient education articles or videos as the response and they can’t answer questions simply. This makes users unable to get an instant and precise answer.
Therefore, to solve the above problems, a health-preserving suggestion extraction model is proposed in this study to extract the health-preserving suggestions from online medical consultation dataset and build the disease knowledge-graph. Based on the knowledge-graph, we propose a health-preserving suggestion chatbot that can understand natural language and provide suggestions for users.
In the experiment, the increasing of entities will decrease as the number of data increases by enhancing the suggestion lexicon. And the accuracy of the model reaches 95.7% in the testing data.

[1] Drift, “The 2018 state of chatbot,” 2018.
[2] GRAND VIEW RESEARCH, “Chatbot Market Analysis By End User, By Application/Business Model, By Type, By Product Landscape, By Vertical, By Region (North America, Europe, APAC, MEA), And Segment Forecasts, 2018 - 2025.”
[3] MEDIOT, “醫生創業打造「醫療版Uber」！鳴醫要讓專業人士不因地點受限｜數位時代.”
[4] N. D.Council, 103-year personal/family digital opportunity survey report. 2014.
[5] Naturali, “AI+医疗：聊天机器人的案例和创新点探讨 | 人人都是产品经理.” [Online]. Available: http://www.woshipm.com/ai/1681965.html. [Accessed: 14-Jul-2019].
[6] 邱國豪, “Based on Intent Frame and Sentiment path to build a semantic understanding system,” 2016.
[7] M.Hu andB.Liu, “Mining and Summarizing Customer Reviews,” 2004.
[8] M.Hu andB.Liu, “Mining Opinion Features in Customer Reviews,” 2004.
[9] A.-M.Popescu andO.Etzioni, “Extracting Product Features and Opinions from Reviews.”
[10] A.Esuli andF.Sebastiani, “SENTIWORDNET: A Publicly Available Lexical Resource for Opinion Mining.”
[11] W.-Y.Ma andY.-Y.Shih, “Extended HowNet 2.0-An Entity-Relation Common-Sense Representation Model,” 2018.
[12] “E-HowNet.” [Online]. Available: http://ehownet.iis.sinica.edu.tw/.
[13] “RASA.” [Online]. Available: https://rasa.com/.
[14] T.Bocklisch, J.Faulkner, N.Pawlowski, andA.Nichol, “Rasa : Open Source Language Understanding and Dialogue Management,” pp. 1–9, 2017.
[15] “Dialogflow.” [Online]. Available: https://dialogflow.com/.
[16] C. ;Yang, K. H.-Y. ;Lin, andH.-H.Chen, “Emotion Classification Using Web Blog Corpora Changhua,” Proc. IEEE/WIC/ACM Int. Conf. Web Intell. WI 2007, pp. 26–34, 2007.
[17] K.Liu, Q.Hu, J.Liu, andC.Xing, “Named entity recognition in Chinese electronic medical records based on CRF,” Proc. - 2017 14th Web Inf. Syst. Appl. Conf. WISA 2017, vol. 2018-Janua, pp. 105–110, 2018.
[18] “快速问医生有问必答网.” [Online]. Available: https://www.120ask.com/. [Accessed: 15-Jul-2019].
[19] “5151線上健康照護網.” [Online]. Available: http://www.5151.tw/.
[20] “Traditional Chinese Medicine.” [Online]. Available: https://21chineseculture.com/thoughts-about-health-preserving/.
[21] 佚名, 黃帝內經. .
[22] Food and Drug Administration, “食品營養成分資料庫(新版).” [Online]. Available: https://consumer.fda.gov.tw//Food/TFND.aspx?nodeID=178&p=1.
[23] Wikipedia, “體育運動列表.” [Online]. Available: https://zh.wikipedia.org/wiki/體育運動列表. [Accessed: 15-Jul-2019].
[24] Z.Dong, Q.Dong, andC.Hao, “HowNet and Its Computation of Meaning,” 2010.
[25] C. J.Lin andH. T.Chang, “Dimentional sentiment analysis by synsets and sense definitions,” Proc. 2016 Int. Conf. Asian Lang. Process. IALP 2016, pp. 332–335, 2017.
[26] “GitHub - fxsjy/jieba: 结巴中文分词.” [Online]. Available: https://github.com/fxsjy/jieba.
[27] 蔡佳純, “Summary Generation for Chinese Patient Complaint based on Medical Entity Recognition and Medical Terminology Mapping,” 2019.
[28] D.Nadeau andS.Sekine, “A survey of named entity recognition and classification.”
[29] C. S.Yang, N.Suh, andA.-N. T.Kong, “Does Vitamin E Prevent or Promote Cancer?,” 2012.
[30] M. of H. and W.Health Promotion Administration, “2015 Taiwan Longitudinal Study on Aging Survey Report,” no. 13, 2015.
[31] “Measuring Performance: The Confusion Matrix – Glass Box Medicine.” [Online]. Available: https://glassboxmedicine.com/2019/02/17/measuring-performance-the-confusion-matrix/.