以虛擬實境為基之智慧變壓器工程諮詢機器人發展與設計

Virtual Reality Enabled Intelligent Transformer Engineering Consultation Chatbot Development and Design

洪楠竣(Nan-Jun Hong)

指導教授 ： 張瑞芬 ； 張力元

國立清華大學/工學院/工業工程與工程管理學系/碩士(2020年)

摘要

日趨增多之網路訊息使得使用者很難快速的從搜尋引擎回傳的大量頁面中找到真正所需內容。具有自然語言處理與理解能力的聊天機器人(chatbot)提供了一個自動問答之媒介，與搜尋引擎不同，它能夠直接回答使用者所需的答案而不是相關的網頁，具有快速高效的優勢。本研究提出了一個以虛擬實境(VR)為基之工程諮詢機器人之系統架構與設計方法，此檢索式聊天機器人(retrieval-based chatbot)蒐集了約1262筆國內外大型機電工程製造商之顧客常問問題(FAQs)與約20個國際標準文件來建構聊天機器人之知識庫，並以維基百科語料庫、相關文獻與工程文件訓練詞向量模型，計算使用者與問答庫問句之相似度來進行資訊檢索以提供適當且完整的回答。此外本研究進一步將聊天機器人與虛擬實境(VR)進行系統整合，結合了自動問答系統與沉浸式科技之高度即時互動性優勢，使得使用者能夠在與系統進行工程諮詢時獲得文字答覆的同時也能與相關產品之三維(3D)模型進行互動。基於提出之系統架構與設計方法本研究建構一電力變壓器製造商提供給顧客之工程諮詢機器人作為系統功能示範，此外也進行問答準確性測試以驗證系統成效與知識庫完整度其結果顯示問答準確度達85 %並與現今常見之語音助理進行比較探討此系統在變壓器工程知識領域之優勢。此系統目的為在虛擬實境環境中提供顧客(例如大型電廠或需大型變壓器之廠商)進行電力變壓器採購作業之相關工程問答諮詢、產出相關的物料清單(BOM)並提供暫定成本資訊，主要虛擬實境系統功能介面著重在呈現電力變壓器及其零組件設計、組合、拆解、客製化等沉浸式互動環境。

關鍵字

聊天機器人 ； 虛擬實境 ； 智慧製造 ； 電力變壓器 ； 自然語言處理

並列摘要

Increasing amount of data and information online make it difficult for users to quickly find what they really need from a large number of web pages returned by search engines. A chatbot with natural language processing and understanding capabilities can provide an interface for automatic question answering. Unlike search engines, it can directly provide answer for user instead of related web pages. This research proposes system framework and design of Virtual Reality (VR) enabled engineering consultation chatbot. The proposed retrieval-based chatbot collects about 1,262 FAQs from transformer manufacturers and about 20 international standard files to build a chatbot knowledge base, and Wikipedia corpus and engineering documents are used to train a word embedding model used in information retrieval to calculate the similarity between the user’s question and the questions in the database to provide appropriate answers. The research integrates chatbot with Virtual Reality (VR), combining advantages of them to enable users conduct engineering consultation and interact with 3D models in real time. Based on the proposed system framework and design, the research constructs a consultation chatbot for power transformer manufacturers as an example and conduct test of question answering accuracy to verify the effectiveness of system. The result shows the answering accuracy is about 85% and is compared with other voice assistants to discuss the advantages of the proposed system in the field of transformer engineering knowledge. The system aims to provide real-time engineering consultation, bill of materials (BOM) for cost estimation, and the main system functions focuses on presenting the immersive interactive environment of power transformers and the components design, disassembly / assembly, customization, etc.

並列關鍵字

Chatbot ； Virtual Reality ； Smart manufacturing ； Power transformer ； Natural language processing

參考文獻

1. Abdul-Kader, S. A., & Woods, J., “Survey on Chatbot Design Techniques in Speech Conversation Systems,” International Journal of Advanced Computer Science & Applications, 6 (7), pp. 72-80, 2015.

2. Adly, A. A., “Computation of Inrush Current Forces on Transformer Windings,” IEEE Transactions on Magnetics, 37 (4), pp. 2855-2857, 2001.

3. Alvarez, J. E., “A Review of Word Embedding & Document Similarity Algorithms Applied to Academic Text,” University of Freiburg,” 2017.

4. Atoum, I., “Scaled Pearson’s Correlation Coefficient for Evaluating Text Similarity Measures, Modern Applied Science, 13 (10), pp.26-38, 2019.

5. Bamodu, O., & Ye, X., “Virtual Reality & Virtual Reality System Components,” Proceedings of the 2nd International Conference On Systems Engineering & Modeling (ICSEM-13), pp. 1169-1172, 2013.