基于百度热搜新闻词的社会风险事件5W提取研究

当今中国处于经济转型升级的关键时期,社会主要矛盾发生了历史性变化,社会风险事件发生的频率比以往更高,危害社会稳定.将公众在线的搜索和关注数据映射为潜在的社会风险事件,如何有效地自动标注风险事件以及直观、清晰地描述社会风险事件是本文关注的重点.本文尝试定义风险事件的5W框架来结构化的描述社会风险,包括地点(where)、时间(when)、人物(who)、原因(why)和发生内容(what).风险事件的5W抽取可转化为不同的机器学习任务,包括命名实体识别、风险分类以及关键词抽取.依据5W的抽取任务进而探索有效的抽取方法.通过对风险事件5W的自动抽取,将现实中社会风险这种wicked问题转化为结构化问题进行分析,为研究社会风险提供一个新的视角,对政府部门进行舆情分析与风险监测具有重要意义.

A study on 5W extraction for societal risk events based on hot news search words

Current China is at a critical period of economic growth transformation and upgrading with the historical changes in the principal contradiction of Chinese society. Risky events occur more frequently than ever, which has an adverse impact on social stability. On-line community concerns have been mapped into respective societal risks. What we concern is how to automatically recognize and distinctly describe societal risk events. We propose an event 5W extraction framework to describe societal risk events in a structured way. 5W refers to where, who, when, why and what. The task for extracting 5W for risk events is converted into different machine learning tasks, which are:Named entity recognition, risk category classification, and keyword extraction. According to tasks of 5W extraction, we try to explore and develop effective methods. The 5W framework allows to automatically extract structured information from societal risk event (a kind of wicked problem) and provides a new perspective to study societal risk, which is of great significance for social management to conduct public opinion analysis and risk monitoring.