| 摘要: | 近年來，聊天機器人越來越流行，各種聊天機器人平台也陸續出現，如主打NLP的公司：Facebook Wit.ai、Google dialogflow、IBM Watson、微軟LUIS等，到無須寫程式的Manychat、Chatfuel平台。
聊天是人與人基本的互動方式，現今通訊軟體已相當普遍，人人都在使用，聊天機器人可以應用在多方面，如線上客服、教育、娛樂及個人助理等等。可以取代重複性高的客服工作，24小時不間斷的工作，讓客服人員可以處理更複雜的問題，節省客服人力資源成本。
本論文研究主要以無伺服器架構來建置學校客服機器人，以Google apps script為開發語言、以Dialogflow進行語意分析以及使用Firebase database儲存對話記錄、滿意度記錄、意見回饋及建置後台讓各行政單位建立問答集。管理者可以依各項數據來了解客服機器人服務狀況，不斷改善問答集，提升準確度，減低客服人員的負擔。
本論文在使用者整體滿意度評價上，針對52位使用者，邀請他們加入客服機器人實際操作後，實施線上問卷，分析結果，李克特量表平均值達4.06，顯示
使用者對本論文所提客服機器人的服務感到滿意。
;In recent years, Chatbots have become more and more popular, and various Chatbot platforms have appeared one after the other, such as NLP companies like Facebook Wit.ai, Google Dialogflow, IBM Watson, Microsoft LUIS, etc., to the Manychat and Chatfuel platforms without writing programs.
Chat is a basic way of interacting with people. Nowadays communication software is quite common and everyone is using it. Chatbots can be used in many ways, such as online customer service, education, entertainment and personal assistants. They can replace the repetitive customer service work, 24 hours of uninterrupted work, so that customer service staff can handle more complex problems, saving the resource cost of customer service manpower.
This study aims to build a school customer service chatbot with serverless framwork, using Google apps script as the development language, analyzing the semantics with Dialogflow, and using Firebase database to store conversation records, satisfaction records, feedback and build the backend for each administration. The unit establishes a question and answer set. Managers can understand the service status of customer service chatbot according to various data, continuously improve the question and answer set, improve accuracy, and reduce the burden on customer service personnel.
In this paper, in the overall customer satisfaction evaluation, 52 users were invited to join the customer service chatbot to implement the online questionnaire, and the results were analyzed. The average value of the Likert scale was 4.06.The user is agreed with the service of the customer service chatbot mentioned in this paper. |
