化工突发事件信息抽取方法的研究
发布时间：2019-02-17 07:46
【摘要】：当今,化工突发事件频发,给人们的生产生活和国家发展造成了巨大的损失和影响,因此对化工突发事件进行有效的监管显得尤为重要。新闻报道是人们了解化工突发事件的一个主要来源,通过对大量新闻报道进行抽取,得到其所包含的相关信息即事故的发生时间、发生地点、事故类型、所涉及的化学品、事故原因、事故结果、事故善后信息,这些信息可以帮助决策者迅速且全面地掌握每一起化工突发事件所暴露的问题,为他们进行科学决策、准确预防从而实现有效监管提供了数据的支持。为了有效地抽取事件的上述信息并应用于化工突发事件信息管理上,本文做了以下的工作:1.通过搜集、整理并阅读大量的化工突发事件新闻报道,统计、分析了新闻报道中,所需抽取的信息的特点,使用模式匹配的方法为每一类信息构建了模式规则库,并为所要抽取的信息制定了相应的抽取算法,进而实现对化工突发事件信息的抽取。2.本文利用最大熵依存句法分析和TextRank关键字提取两种机器学习算法,根据用户对信息抽取结果的评价来自动生成新的规则模式,从而实现了事故信息抽取规则的自学习,拓宽了模式规则库,提高了下一次信息抽取的准确率。3.基于上述算法并且结合中国石化安全工程研究院的业务需求,本文设计开发了化工突发事件信息抽取系统,实现了对化工突发事件信息的抽取与模式规则的自动生成。通过用户的反馈以及通过大量的实验表明,该信息抽取方法有较高的准确率,抽取的效果较为的理想。
[Abstract]:Nowadays, the frequent occurrence of chemical emergencies has caused great loss and influence to people's production and life and the development of our country. Therefore, it is very important to supervise chemical emergencies effectively. News reports are one of the main sources for people to understand chemical emergencies. By extracting a large number of news reports, we can get the relevant information, that is, the time, place, type, chemical involved of the accident, that is, the time when the accident occurred, the location of the accident, the type of accident, and the chemical involved. The cause of the accident, the result of the accident, the information about the aftermath of the accident, which can help decision makers quickly and comprehensively grasp the problems exposed by every chemical accident, and make scientific decisions for them. Accurate prevention to achieve effective regulation provides data support. In order to extract the above information effectively and apply it to the management of chemical emergency information, this paper has done the following work: 1. By collecting, sorting and reading a large number of news reports on chemical emergencies, this paper analyzes the characteristics of the information that needs to be extracted from the news reports, and constructs a pattern rule base for each type of information by using the method of pattern matching. A corresponding extraction algorithm is developed for the information to be extracted, and then the information of chemical emergencies is extracted. 2. 2. In this paper, two machine learning algorithms, maximum Entropy dependent Syntax Analysis and TextRank keyword extraction, are used to automatically generate new rule patterns according to users' evaluation of information extraction results, thus self-learning of accident information extraction rules is realized. It broadens the pattern rule base and improves the accuracy of the next information extraction. 3. 3. Based on the above algorithm and the business requirements of Sinopec Institute of Safety Engineering, this paper designs and develops a chemical emergency information extraction system, which realizes the extraction of chemical emergency information and the automatic generation of pattern rules. Through user feedback and a large number of experiments, it is shown that this information extraction method has a high accuracy and the effect of extraction is more ideal.
【学位授予单位】：青岛科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.1
本文编号：2424919
[Abstract]:Nowadays, the frequent occurrence of chemical emergencies has caused great loss and influence to people's production and life and the development of our country. Therefore, it is very important to supervise chemical emergencies effectively. News reports are one of the main sources for people to understand chemical emergencies. By extracting a large number of news reports, we can get the relevant information, that is, the time, place, type, chemical involved of the accident, that is, the time when the accident occurred, the location of the accident, the type of accident, and the chemical involved. The cause of the accident, the result of the accident, the information about the aftermath of the accident, which can help decision makers quickly and comprehensively grasp the problems exposed by every chemical accident, and make scientific decisions for them. Accurate prevention to achieve effective regulation provides data support. In order to extract the above information effectively and apply it to the management of chemical emergency information, this paper has done the following work: 1. By collecting, sorting and reading a large number of news reports on chemical emergencies, this paper analyzes the characteristics of the information that needs to be extracted from the news reports, and constructs a pattern rule base for each type of information by using the method of pattern matching. A corresponding extraction algorithm is developed for the information to be extracted, and then the information of chemical emergencies is extracted. 2. 2. In this paper, two machine learning algorithms, maximum Entropy dependent Syntax Analysis and TextRank keyword extraction, are used to automatically generate new rule patterns according to users' evaluation of information extraction results, thus self-learning of accident information extraction rules is realized. It broadens the pattern rule base and improves the accuracy of the next information extraction. 3. 3. Based on the above algorithm and the business requirements of Sinopec Institute of Safety Engineering, this paper designs and develops a chemical emergency information extraction system, which realizes the extraction of chemical emergency information and the automatic generation of pattern rules. Through user feedback and a large number of experiments, it is shown that this information extraction method has a high accuracy and the effect of extraction is more ideal.
【学位授予单位】：青岛科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.1
【参考文献】
相关期刊论文 前10条
1 李志红;;石油化工企业安全事故原因分析及对策研究[J];石油化工安全环保技术;2013年06期
2 夏天;;词语位置加权TextRank的关键词抽取研究[J];现代图书情报技术;2013年09期
3 李淼;杜明晶;苗放;;网页设计中Bootstrap CSS框架的应用与拓展[J];电子技术与软件工程;2013年17期
4 宋子辉;;自然语言理解的中文地址匹配算法[J];遥感学报;2013年04期
5 孙波;刘丹;;基于MySQL的嵌入式移动数据库系统研究[J];河南机电高等专科学校学报;2012年03期
6 仲秋雁;郭艳敏;王宁;薛慧芳;崔丽;王延章;;基于知识元的非常规突发事件情景模型研究[J];情报科学;2012年01期
7 奉国和;郑伟;;国内中文自动分词技术研究综述[J];图书情报工作;2011年02期
8 蒋德良;;基于规则匹配的突发事件结果信息抽取研究[J];计算机工程与设计;2010年14期
9 孔芳;周国栋;朱巧明;钱培德;;指代消解综述[J];计算机工程;2010年08期
10 辛霄;范士喜;王轩;王晓龙;;基于最大熵的依存句法分析[J];中文信息学报;2009年02期
相关博士学位论文 前2条
1 谭红叶;中文事件抽取关键技术研究[D];哈尔滨工业大学;2008年
2 杨尔弘;突发事件信息提取研究[D];北京语言大学;2005年
相关硕士学位论文 前1条
1 吴平博;基于事件框架的主题相关文档智能检索的初步研究[D];清华大学;2004年
，本文编号：2424919
本文链接：https://www.wllwen.com/kejilunwen/ruanjiangongchenglunwen/2424919.html
