運用Web文字探勘方法來建構領域知識庫：以客戶服務與聊天機器人為研究案例

Applying Web Text Mining Approaches to Build Domain Knowledge Base: Case Studies on Customer Services and Chatbots

顏皓民(Hao-Min Yan)

指導教授 ： 林宣華

國立暨南國際大學/科技學院/資訊工程學系/博士(2017年)

摘要

企業客服模式從傳統電話客服轉為近期行動裝置Apps和客服聊天機器人 (Chatbot)。客服聊天機器人能自動且透明化地保留客戶聊天記錄 (Chat Logs) 並建構大數據應用，進而改善【客戶關係管理 (Customer Relation Management, CRM)】讓決策更有效率。然而，開發客服聊天機器人需專家和員工的配合，透過分析企業資料來建置機器人的知識，其過程需花費大量的人力與時間。因此，我們提出的自動化領域知識建構系統 (Automatic Domain Knowledge Construction System, ADKC) 能協助使用者自動化建構領域知識庫。只要給定領域名稱和幾個重要的關鍵字，系統便會透過知識蒐集 (Knowledge Collection, KC) 階段從網路上蒐集相關資料並擷取關鍵字，再將資料做領域與非領域的分類。接著，知識探索 (Knowledge Exploration, KE) 階段從領域資料中擷取概念並定義主題，再基於主題由上而下分層建立領域知識樹。根據過程擷取到的領域關鍵字，系統能自動且反覆地運行此兩個階段的模組來持續學習領域知識。本論文以「食安 (Food Safety)」和「王品 (WowPrime)」做為研究案例，探討知識樹為核心所建立的知識庫，展示系統成果，並應用於智慧客服機器人。

關鍵字

文本探勘 ； 知識庫 ； 資料分類 ； 概念擷取 ； 知識樹 ； 智慧型客服

並列摘要

With the development of digital economy, services from business to customer (B2C) are moving from traditional phone contact with manpower to mobile Apps through intelligent Chatbots (Chat Robots). Chatbots automatically and transparently keep customer logs so that building big data applications based on chat logs makes improving customer relationship management (CRM) for decision-making more effective and efficient. However, building intelligent customer services like Chatbots have to spend expensive efforts to integrate domain experts and staffs to curate cooperation knowledge bases. Therefore, we propose Automatic Domain Knowledge Construction (ADKC) system to automatically build domain knowledge bases. Given domain name and optional several domain-related keywords, ADKC first activates Knowledge Collection (KC) subsystem that collects domain-related data from the Web, extracts domain candidate keywords and classify data into domain and non-domain categories. Then, Knowledge Exploration (KE) subsystem gradually refines candidate keywords into important domain concepts and topics so that the domain knowledge tree can be hierarchically built based on topic-based top-down approach. ADKC iteratively runs modules of both subsystems to automatically and continuously learn domain knowledge based on criteria of measuring qualities of domain keywords. Based on knowledge bases automatically constructed by ADKC, we verify results according to case studies on domains: “Food Safety” and “WowPrime”. Finally, the knowledge bases of WowPrime is also applied to build Chatbot applications.

並列關鍵字

text mining ； knowledge base ； data classification ； concept extraction ； knowledge tree ； intelligent customer service

參考文獻

[1] Agrawal, R., Imieliński, T., & Swami, A. (1993, June). Mining association rules between sets of items in large databases. In Acm sigmod record (Vol. 22, No. 2, pp. 207-216). ACM.

[2] Bickel, S., & Scheffer, T. (2004, November). Multi-view clustering. In ICDM (Vol. 4, pp. 19-26).

[3] Chakrabarti, S., Van den Berg, M., & Dom, B. (1999). Focused crawling: a new approach to topic-specific Web resource discovery. Computer networks, 31(11), 1623-1640.

[4] Chien, L. F. (1997, July). PAT-tree-based keyword extraction for Chinese information retrieval. In ACM SIGIR Forum (Vol. 31, No. SI, pp. 50-58). ACM.

[5] Feng, D., Shaw, E., Kim, J., & Hovy, E. (2006, January). An intelligent discussion-bot for answering student queries in threaded discussions. In Proceedings of the 11th international conference on Intelligent user interfaces (pp. 171-177). ACM.