Please use this identifier to cite or link to this item:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/21676
Title: | 探究自然語言處理與聊天機器人於營建管理資料查詢處理之應用 Exploring the Application of NLP and Chatbot for Construction Data Searching and Processing |
Authors: | Tsai-Ning Chen 陳采寧 |
Advisor: | 陳柏翰(Po-Han Chen) |
Keyword: | 自然語言處理,聊天機器人,BERT模型,人工智慧,營建管理, NLP,chatbot,BERT,AI,construction management, |
Publication Year : | 2020 |
Degree: | 碩士 |
Abstract: | 近年來人工智能(Artificial Intelligence)發展快速,綜觀AI在許多領域的傑出表現,我們可以相信幾乎在任何未來的產業裡「自動化」、「智能化」是勢在必行的趨勢,幾乎每個產業都受益於人工智能而發展得更加快速。然而相對於電子科技業,土木界於此方面的應用可以更多更廣。觀察到現今土木產業的合約、規範等文件資料量龐大,搜尋所需訊息耗費時間長,加上目前大多中小型企業公的方式進行資料的整理查詢,因此希望能利用AI相關子領域:自然語言處理,開發一套問答系統提供土木人員快速搜尋想知道的訊息。此篇論文應用Google於2018年發佈的自然語言處理(NLP)技術Bidirectional Encoder Representations from Transformers (BERT)模型去處理各種文件資訊,增強現有的自然語言模型文件訊息擷取能力,並希望解決因資訊取得不夠快速所造成的許多問題,如工地現場溝通糾紛、效率低落等等。 此篇研究從以往常用的關鍵字搜尋改成語意相似度查詢模式,並配合通訊軟體Line讓使用者能更便利的在常用的通訊軟體介面中使用更強大的自然語言模型去搜尋所需要的資訊,此份研究做出一個聊天機器人為土木人提供一套即時的QA系統,並加入能使模型更理解口語化問答的語料集,可提供使用如工地主任以日常口語詢問的方式做:(1)不同階段、不同實施項目的相關法規(2)案件合約內容(3)材料規範 (4)測試標準 等等查詢。透過精確度測試機制與口語化測試機制,分析實際操作時所遇之問答情況並找出最佳的使用模式,使此系統更符合真實問答系統的操作需求。 Recently, the development of AI has increased dramatically. In overall picture, the performance of AI in various fields has been tremendous. It is very obvious and persuasive that “Automation” and “Intellectualization” is an inevitable but beneficial trend in almost every industry. Comparing to Electronic Technology Industry, the Civil Engineering Industry certainly has more space and unrevealed potential for the relevant implement. The large consumption of time due to the massive amount of regulations and related contracts in the current civil engineering industry has been noticed, therefore, a professional system is developed with AI relevant technology (Natural Language Processing) specifically for civil engineers to efficiently search for relevant required-information. This dissertation adopted the Bidirectional Encoder Representations from Transformers (BERT) which is presented by Google in 2018 to process all categories of document materials、enhance the automated data capture capability of the current language model; moreover, to solve the problems caused by inefficient information acquirement such as arguments between customers and workers or schedule delays. The dissertation aims to design a new system with a combination of BERT and a communication application: Line to produce an immediately-response QA system. In addition, the system has been improved from the traditional key word searching to similar semantics searching; furthermore, the system was provided with oral language database in order to analyze more oral language. The QA system will be able provide the related information of the following categories: 1. The regulations of the different stage、construction items 2. The content of the contract for the designated case 3. The regulations of the designated materials while using in the construction sites 4. The spec or standard for different monitor stage or test Through accuracy and oral language assessment, the system is expected to provide the most suitable answer to the actual situation. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/21676 |
DOI: | 10.6342/NTU202100056 |
Fulltext Rights: | 未授權 |
Appears in Collections: | 土木工程學系 |
Files in This Item:
File | Size | Format | |
---|---|---|---|
U0001-1301202115513200.pdf Restricted Access | 5.51 MB | Adobe PDF |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.