113

4,235 千元

化學物質科學研究-化學物質科學研究

2024/04/12

2024/11/30

曾宇鳳

化學署綜合規劃組

李昱宣

國立臺灣大學

科技計畫

綠色化學、安全替代、危害資料庫

Green chemistry, Safer alternatives, Database for hazardous chemical

郭天爵

02-33664888#403

cot@cmdm.csie.ntu.edu.tw

本計畫旨在發展「高風險化學物質之安全替代化學物質搜尋、評估及篩選系統（Searching, Assessment, and Screening System for Safer Alternative Chemicals, SAS）」的診斷模組系統。該系統將基於模組架構進行優化，其功能包括對化學物質危害進行診斷、建立安全替代建議清單，以便促進 化學物質的源頭管理。 113 年計畫配合國際安全替代評估策略及運算邏輯、化學物質管理法規、危害物質或管制清單，持續優化SAS診斷模組系 統。團隊增強替代評估運算邏輯與加入ChemSec Marketplace與美國國家環境環保局 ChemView 擴充化學品產業別資訊，藉此可增強產業別參數，連 結現有不同產業中安全替代化學品，增強個別化合物替代建議。藉由人工檢查現有國際清單，更新爬蟲程式，擴充安全替代化學物質資料庫與關聯 資料庫，以提升其應用性，為了強化使用者對於判讀診斷模組資訊，團隊加入診斷模組教學文件，說明四種診斷模組原理，以及所提供之資訊。團 隊亦強化搜尋功能、並將中文名稱與俗名（如農藥名）共 1,790 筆中文名稱加入資料庫提升業者使用便利性。此外，本計畫以AI大型語言模型基於 化學品毒性資訊與安全替代物資訊，藉由 Meta Llama3 模型與 NVIDIA NeMo 程式柵欄實作對談機器人，由使用者與 AI 系統對談取得所需之資 料，並且忽略無關於化學品與危害問題，以使 SAS 診斷模組系統更加完善、提升使用者體驗。

綠色化學、安全替代、危害資料庫

113化學署_成果報告v8.pdf

7MB

2025/2/16 上午 12:00:00

This project aims to develop the Diagnostic Module System of the "Searching, Assessment, and Screening System for Safer Alternative Chemicals (SAS)" for high-risk chemicals. The system will be optimized based on a modular framework, with functionalities including hazard diagnosis of chemicals and the establishment of a list of recommended safer alternatives to promote source management of chemicals. In 2024, the team, will align with international assessment strategies for safer alternatives, computational logic, chemical management regulations, and hazard or control lists, continuously optimizing the SAS diagnostic module system. The team will enhance the alternative assessment computational logic and integrate ChemSec Marketplace and the US EPA ChemView to expand industry-specific chemical information, thereby strengthening industry-specific parameters and linking existing safer alternative chemicals across different industries. This will enhance individual compound substitution recommendations. By manually reviewing existing international lists, updating crawler programs, and expanding the safer alternative chemicals database and related databases, the applicability of the system will be improved. To enhance user interpretation of diagnostic module information, the team will include tutorial documents in the diagnostic module, explaining the principles of the four diagnostic modules and the provided information. The team will also enhance the search functionality and add 1,790 Chinese names and common names (e.g., pesticide names) to the database to improve usability for industry practitioners. Additionally, the project will use AI large language models, based on chemical toxicity information and safer alternative information, to implement a conversational chatbot using the Meta Llama3 model and NVIDIA NeMo framework. This will allow users to acquire necessary information through interactions with the AI system while ignoring irrelevant chemical and hazard issues, thereby improving the SAS diagnostic module system and enhancing the user experience.

Green chemistry, Safer alternatives, Database for hazardous chemical