跳到主要內容 :::

年度	112	專案性質	非實驗性質
專案類別	研究專案	研究主題	自訂
申請機構	東海大學	申請系所	環境科學與工程系
專案主持人	陳鶴文	職等/職稱	特聘教授
專案中文名稱	土壤及地下水污染管理與政策輔助平台開發：生成式人工智慧技術之應用
中文關鍵字	土壤與地下水汙染控制與品質管理,生成式人工智慧技術
專案英文名稱	Development of Soil and Groundwater Pollution Management and Policy Support Platform: Application of Generative Artificial Intelligence Technology
英文關鍵字	Control and Quality Management of Soil and Groundwater Pollution, Generative Artificial Intelligence Technology
執行金額	1,024,000元
執行期間	2023/12/15 至 2024/11/29
計畫中文摘要	生成式人工智慧技術可以協助使用者從大量文本資料中，快速找出使用者需要的資訊並組織成使用者可以理解的文句，使知識的分享與應用更加有效。資料與資訊通常需要經過學習以及價值認知才能轉化成有用的知識。如何將環境部已累積的大量研究報告、法規、案例資料與技術手冊等文本資料，透過收集、組織、管理、分享與應用的步驟將環境部內部的知識與經驗分享出去，並支持環境部內部進行各項的管理和決策程序是本計畫的目標。環境問題是由各種不同的事件和案例所組成，若能從過去的事件與案例中累積經驗，便可避免未來土壤與地下水之污染事件發生，且更有效率的解決新污染事件。因此，本研究以土壤與地下水管理之科研報告主題為例，建置土壤與地下水知識管理與應用平台，目前以「現地奈米整治技術」、「生物免疫法」主題來進行平台建置，該平台以完成關鍵字詞萃取、知識關聯網絡、主題式知識本體網絡及大型語言模型，從結果發現主題式知識本體網絡可有效提升大型語言模型的問答精準度，減低答非所問的情形，並且可以提供相對應的參考文獻資料，讓使用者可以佐證此回答是否正確。完成此平台後，本研究邀請政府機構人員進行施測，其測試結果皆針對報告書來進行回答，並無出現答非所問的狀態。由此可知，此平台的建置可以提供給相關部門未來培訓教育人員或者在思考策略時，做為一個找出策略的科學化工具。
計畫英文摘要	Generative artificial intelligence technology can help users quickly find the information they need from large amounts of textual data and organize it into sentences that users can understand, making knowledge sharing and application more effective. Data and information usually need to undergo learning and value recognition before being transformed into useful knowledge. The goal of this project is to share the accumulated knowledge and experience within the Ministry of Environment by collecting, organizing, managing, sharing, and applying large volumes of research reports, regulations, case studies, and technical manuals, while supporting various management and decision-making processes within the Ministry. Environmental issues consist of various events and cases. If experiences can be accumulated from past events and cases, future soil and groundwater pollution incidents can be prevented, and new pollution events can be resolved more efficiently. Therefore, this research focuses on the topic of scientific reports related to soil and groundwater management to establish a knowledge management and application platform. The platform construction currently centers on "in-situ nano-remediation technology" and "bio-immunization." The platform incorporates keyword extraction, knowledge association networks, topic-based knowledge ontology networks, and large language models. Results show that the topic-based knowledge ontology network can effectively improve the accuracy of large language model responses, reduce irrelevant answers, and provide corresponding reference materials to allow users to verify the correctness of the responses. After completing the platform, this research invited government officials to conduct testing. The test results indicated that the responses were all based on the content of the reports, with no cases of irrelevant answers. This demonstrates that the platform's construction can serve as a scientific tool for relevant departments in future training and education or in strategic planning to identify strategies.