An Enjoyable QA-pair Spoken Dialogue System with the following charisteristcs is presented:
1. Fast: A two-stage fast comparison method for similar sentences.
2. Enjoyable: Synonym substitution, syntactic structure transform and template matching methods.
3. Completion: When the feedback sentence is not found, use deep learning to search for the article to respond appropriately.

本技術的特點主要分為三個部分： 
1.使用兩階段相似句快速比對方法，此方法可因應不同領域的應用更換QA-pair，並且省去了深度學習所耗費訓練的時間，藉此可達到便捷性。
2.透過詞嵌入模型(Word-embedding)搭配分析樹(Parse tree)與槽填充(Slot-filling)使得回應句更具趣味性。
3.結合深度學習中的注意力架構(Attention Architecture)與長短期記憶模型(LSTM)，讓本對話系統賦予閱讀文章的能力，這可解決QA-pair不夠完整時，依然能夠回饋答案給使用者，藉而提升本對話系統之完整度。
以上技術再配合我們蒐集的台南美食問答語料庫與台南美食文章語料庫，使本系統讓使用者得以輕鬆地詢問台南美食。本系統MOS整體評價為4.4分。

The proposed system can be customized according to the needs of users. By changing the training text, it can be adapted to each field and decreases human resource requirements. The field applications include recreational services, smart homes and smart medical care.