近年來因社群媒體與即時通訊軟體的普及，使得行動訊務量快速增加，透過手機聊天已成為人與人之間最簡單且最直接的互動方式，因此造就聊天機器人此種新的服務型態。透過與機器人簡單的對話來獲得資訊或是藉由機器人將工作既定的流程簡化來降低資源成本，已成為業界研究的熱門主題。聊天機器人（Chatbot）是經由對話或文字進行交談的電腦程式，藉由自然語言的學習與訓練能夠模擬人類對話，給予適當的回饋資訊。
本論文研究主要以Python為程式開發語言，利用開源的網頁服務框架Django、微軟QnA Maker與LUIS自然語意辨識服務來設計與建構聊天機器人。以往建立聊天機器人必須熟知自然語言和海量語料資料的訓練才能得到有效的成果，此系統重點著重在節省資料訓練的時間及快速部署校園諮詢機器人客服系統，利用現有的校園FAQ網站與問題集，透過自動化檢索與訓練來實作機器人客服，強化既有工作流程。
本研究除實作出機器人客服外，另提出幾項管理機制，其中有對話歷史紀錄機制、提問量紀錄、服務品質回饋、線上轉接真人與問卷評分調查等。透過簡易管理與訓練、視覺化的圖形化監測服務與數據分析，對管理者而言可以隨時掌握機器人客服與使用者的互動情況，確保整體服務的連貫。最重要的是可以分擔客服的工作量，累積客服經驗，持續不斷的完善問題庫，更讓整題服務滿意度提升12.63%。

Due to popularize of social networks and instant messaging applications, mobile usage has grown at an unprecedented pace and has become the most efficient way for people to connect. As a result, sparked a new type of service through automated conversational agents, Chatbot. Lowering costs by disseminating information through predesigned processes has become a hot industry research topic. A chatbot is an automated response system through texts or voice messaging, achieved by artificial learning to simulate human communication.
This research based on Python for program development utilizes open source web framework Django, Microsoft’s QnA Maker and LUIS for text recognition to design chatbot. Previous chatbot developments require proficient natural language cognition and training through massive language information to reach effective results. This system, however, focuses on minimizing information training time to achieve quicker implementation of in-school information chatbot system; by gathering school’s FAQ pages and threads, automated collection and training will strengthen chatbot’s work processes for effective services.
Other than developing chatbot services, this research also suggests managing protocols, including chat logs, inquiring session logs, service quality feedbacks, transference to person and rating feedbacks. Through simplified management and training, visualized monitoring tools, data analysis, backend operator can grasp the effectiveness of interactions in real-time, making sure quality fits its entirety. Most importantly, alleviate servicing staff’s workload, accumulate experience and data, resulting in improvement of the service quality that reaches to 12.63%.

中文摘要 i
Abstract ii
致謝 iii
目錄 iv
圖目錄 vi
表目錄 viii
第一章緒論 1
1-1研究動機 2
1-2研究目的 2
1-3章節架構 3
第二章 背景知識與相關研究 4
2-1 人工智慧 4
2-2 聊天機器人發展與理論技術 7
2-3 聊天機器人案例探究 14
2-3-1 Eliza對話機器人 14
2-3-2 Alice對話機器人 15
2-3-3 Siri對話機器人 15
2-3-4 微軟小冰機器人 16
2-4 電商網站聊天機器人 17
2-5 文獻探討 18
第三章 研究方法 21
3-1系統架構與設計 21
3-2系統開發工具與平台 25
3-2-1 Django與 Python 25
3-2-2 Microsoft QnA Maker 27
3-2-3 Microsoft LUIS 29
3-2-4 Line＠Manager 29
3-2-5 Line Message API 29
3-2-6 Heroku 31
3-3系統模組 32
3-3-1 知識集與資料訓練模組 33
3-3-2 語意分析模組 34
3-3-3 對話與客製選單模組 35
3-3-4 線上真人客服服務模組 36
3-3-5 對話紀錄模組 37
3-3-6 機器人提問量紀錄模組 38
3-3-7 機器人客服評價紀錄模組 39
3-3-8 機器人客服評價模組 40
3-3-9機器人客服問卷模組 41
3-4系統運作服務 42
3-4-1 機器人客服服務流程 42
3-4-2 機器人客服資料訓練流程 44
3-4-3 機器人客服後台系統說明 46
第四章 實驗與討論 47
4-1 實驗環境 47
4-2 實驗數據統計 48
4-2-1機器人客服問題解決率分析 48
4-2-2機器人客服選單使用比率分析 50
4-2-3機器人客服好友趨勢分析 51
4-2-4機器人客服假日使用率分析 52
4-2-5機器人客服問答滿意度評價 53
4-2-6機器人客服有無選單滿意度分析 54
4-3實驗討論 55
4-3-1機器人客服與傳統客服服務比較 56
4-3-2 機器人客服服務問卷統計 57
第五章結論與未來研究方向 61
5-1結論 61
5-2研究限制 62
5-3未來研究方向 62
參考文獻 64

[1] Cai Shushan, and Xue Xiaodi, ‘‘Artificial Intelligence and Human Intelligence：A view on Human-Computer Competition from the Five-Level Theory of Cognitive Science,’’Journal of Peking University(Philosophy and Social Sciences), Vol53 No.4, Jul 2016.

[2]Wolfgang Bibel, “Artificial Intelligence in a historical perspective,” DOI 10.3233/AIC- 130576, pages 87-102, AI Communications 27,6 Dec 2014

[3] STPI, Chatbot User Case Report,[Online] Available:https://outlook.stpi.narl.org.tw/index/focusnews/detail/351[Accessed: 28-JUN-2018]

[4] A. M. Turing, “Computing Machinery and Intelligence”, Mind, Volume LIX, Issue 236, pages 433–460,1 October 1950

[5]Chen Hongshen, Liu Xiaorui, Yin Dawei, and Tang Jiliang, “A Survey on Dialogue Systems: Recent Advances and New Frontiers,” Data Science and Engineering Lab, Michigan State University, preprint arXiv:1711.01731, Nov 2017

[6] M. Dahiya, “A tool of conversation: chatbot,” International Journal of Computer Sciences and Engineering Vol.5(5), May 2017

[7]張偉男，劉挺，“聊天機器人技術的研究進展”，中國人工智慧學會通訊，第6卷第1期，2016.

[8]W. Joseph, “ELIZA—a computer program for the study of natural language　communication between man and machine,” Communications of the ACM, vol. 9, Issue 1,pp.36-45,1966

[9] X. Tian, F. X. Zhong, and L. Lin, “ALICE Mechanism Analysis and Application Study,” Journal of Computer Applications, vol. 23, No. 9, 2003.

[10]L.Caviglione and W.Mazurczyk,“Understanding Information Hiding in iOS,” Computer,pp.62 – 65,2015


[11]Microsoft,Xiaobing,Wiki,[Online].Available: https://zh.wikipedia.org/wiki/%E5%B0%8F%E5%86%B0,
[Accessed: 28-JUN-2018]

[12] Siddharth Gupta, “An E-Commerce Website based Chatbot,” (IJCSIT) International Journal of Computer Science and Information Technologies, Vol. 6 (2), 2015

[13]Williams, Jason D., Asadi, Kavosh, and Zweig, Geoffrey,
“Hybrid Code Networks: practical and efficient end-to-end dialog control
with supervised and reinforcement learning”, 55th Annual Meeting of the Association for Computational Linguistics, eprint arXiv:1702.03274,2017

[14] Kyo-Joong Oh, Dongguan Lee, Byungoo Ko, and Ho-Jin Choi, “A Chatbot for Psychiatric Counseling in Mental Healthcare Service Based on emotional dialogue Analysis and sentence generation”, IEEE 18th International Conference on Mobile Data Management,pp.371-376,2017

[15]Godson Michael D’silva1, Sanket Thakare2, Shraddha More1, and Jeril Kuriakose1, “Real World Smart Chatbot for Customer Care using a Software as a Service (SaaS) Architecture,” International conference on I-SMAC,pp.658-664,2017

[16]Supratip Ghose, and Jagat Joyti Barua, “Toward the implementation of a Topic-specific Dialogue based Natural Language Chatbot as an Undergraduate Advisor,” International Conference on Informatics, Electronics and Vision (ICIEV),2013

[17]魏彰村，運用爬蟲技術之主題導向即時通訊聊天機器人，設計與實現:以籃球領域諮詢結合LINE APP實作為例，中正大學碩士論文，2017。

[18]Daniel Braun, Adrian Hernandez Mendez, Florian Matthes, and Manfred Langen,‘‘Evaluating Natural Language Understanding Services for conversation question answering system,” Proceedings of the SIGDIAL 2017 Conference, pages 174–185, Saarbru¨cken, Germany, pp.15-17 August 2017.



[19] Bocklisch Tom, Faulkner Joey, Pawlowski Nick and Nichol Alan,‘‘Rasa: Open Source Language Understanding and Dialogue Management,’’ Presented at NIPS Workshop on Conversational AI, eprint arXiv:1712.05181, Dec 2017.

[20]Prototype, [Online] Available: https://en.wikipedia.org/wiki/Prototype

[21]Microsoft QnA Maker, ‘‘QnAMaker.’’ [Online]
Available: https://www.qnamaker.ai [Accessed:28-Jun-2018].


[22]Microsoft LUIS, ‘‘LUIS’’, [Online]. Available: https://www.luis.ai/home
[Accessed:28-Jun-2018].

[23]Django,[Online].Available:https://djangogirlstaipei.gitbooks.io/django-girls-taipei-tutorial.[Accessed:28-Jun-2018].

[24]Python, [Online].Available: https://www.python.org, [Accessed:28-Jun-2018].

[25]Line@, [Online].Available: https://at.line.me/tw,[Accessed:28-Jun-2018].

[26]Line API, [Online].Available: https://developers.line.me/en/docs/messaging-
api/overview,[Accessed:28-Jun-2018].

[27]Heroku cloud platform tutorial,[Online].
Available:http://lee-.github.io/posts/bot/2016/11/deploy-linebot-on-heroku,
[Accessed:28-Jun-2018].

[28] Zachary C. Lipton, John Berkowitz and Charles Elkan,‘‘A Critical Review of Recurrnet Neural Networks for Sequence Learning,’’arxiv:1506.00194[cs.LG],17 Oct 2015.