文献类型：期刊文献

英文题名：A knowledge graph method for hazardous chemical management: Ontology design and entity identification

作者：Zheng, Xue[1];Wang, Bing[1];Zhao, Yunmeng[1];Mao, Shuai[1];Tang, Yang[1]

机构：[1]East China Univ Sci & Technol, Key Lab Adv Control & Optimizat Chem Proc, Minist Educ, Shanghai 200237, Peoples R China

年份：2021

卷号：430

起止页码：104

外文期刊名：NEUROCOMPUTING

收录：;EI(收录号：20205109660558);WOS:【SCI-EXPANDED(收录号:WOS:000617365300010)】；

基金：This work is supported in part by the National Key Research and Development Program of China under Grant 2018YFC0809302, the National Natural Science Foundation of China under Grants 61988101, 61751305, 61673176 and the Programme of Introducing Talents of Discipline to Universities (the 111 Project) under Grant B17017.

语种：英文

外文关键词：Knowledge graph; Ontology; Hazardous chemicals management; Named entity recognition

摘要：Hazardous chemicals are widely used in the production activities of the chemical industry. The risk management of hazardous chemicals is critical to the safety of life and property. Hence, the effective risk management of hazardous chemicals has always been important to the chemical industry. Since a large quantity of knowledge and information of hazardous chemicals is stored in isolated databases, it is challenging to manage hazardous chemicals in an information-rich manner. Herein, we prompt a knowledge graph to overcome the information gap between decentralized databases, which would improve the hazardous chemical management. In the implementation of the knowledge graph, we design an ontology schema of hazardous chemicals management. To facilitate enterprises to master the knowledge in the full lifecycle of hazardous chemicals, including production, transportation, storage, etc., we jointly use data from companies and open data from the public domain of hazardous chemicals to construct the knowledge graph. The named entity recognition task is one of the key tasks in the implementation of the knowledge graph, which is of great significance for extracting entity information from unstructured data, namely the hazardous chemical accidents records. To extract useful information from multi-source data, we adopt the pre-trained BERT-CRF model to conduct named entity recognition for incidents records. The model achieves good results, exhibiting the effectiveness in the task of named entity recognition in the chemical industry. (c) 2020 Elsevier B.V. All rights reserved.