摘要
针对抽取新闻关键信息需要人工构造或训练生成模板的问题,提出了基于标签和分块特征的新闻关键信息自动抽取方法。该方法首先通过计算新闻网页相关特征来定位新闻正文标签块,然后通过编辑距离定位新闻标题标签块,最后根据正文块和标题块定位新闻发布时间和来源标签块,并通过抽取各块的文本获得目标新闻关键信息。在该方法的基础上提出了针对新闻站点的目标新闻自动抽取框架,并用该框架对10个新闻站点的30个新闻栏目进行了新闻抽取。对抽取到的1 597条新闻随机选择了1 000条进行了实验。实验结果表明,该方法对新闻标题、发布时间、来源、正文均表现出良好的抽取效果,且优于实验对比对象。
In view of that issue of news key information extraction require manual construction template or training generation template, provide news key information automatic extraction method based on label and block characteristics. The method first locates the news body block by calculating the relevant characteristics of the text block, then position the news title block by editing distance, and finally locates the press release time and source label block according to the text block and title block, and finally obtains the target news key information by extracting the text of each block. On the basis of this method, propose an automatic news extraction framework for news sites, and uses this framework to extract news from 30 news columns of 10 news sites. A total of 1 597 news data are collected, and 1 000 of them are randomly selected for the experiment. The experimental results show that this method has a good extraction effect on news title, publish time, source and text, and is superior to the comparison objects.
作者
王雪梅
陈兴蜀
王海舟
王文贤
Xue-mei WANG;Xing-shu CHEN;Hai-zhou WANG;Wen-xian WANG(College of Software Engineering, Sichuan University, Chengdu 610065, Sichuan, China;College of Cybersecurity, Sichuan University, Chengdu 610065, Sichuan, China;Cybersecurity Research Institute, Sichuan University, Chengdu 610065, Sichuan, China)
出处
《山东大学学报（理学版）》
CAS
CSCD
北大核心
2019年第3期67-74,共8页
Journal of Shandong University(Natural Science)
基金
国家自然科学基金资助项目(61802270
61802271)
国家"双创"示范基地之变革性技术国际研发转化平台资助项目(C700011)
四川省重点研发项目资金资助(2018G20100)
四川省科技支撑计划资金资助(2016GZ0038)
中央高校基本科研业务费专项资金资助(2017SCU11065)
