摘要
[目的/意义]从事件文本句法特征视角出发,提出一种面向突发自然灾害的网络舆情事件识别方法,使得从小规模数据集中精准的识别事件成为一种可能。[方法/过程]通过数据采集和事件语义标注构造训练集,接着提出了一种面向突发自然灾害网络舆情事件识别的句法特征提取方法,利用句法特征提取方法从训练集中提取事件句法构造事件句法特征库,同时以句法向量的形式表示待测事件文本,最后利用事件句法与待测句法的句法相似度计算识别事件。[结果/结论]以“台风利奇马”事件为例,证明了本研究提出的事件识别方法能够精准地从突发自然灾害网络舆情文本中识别事件,同时通过对照试验证明了在训练集规模较小的情况下,句法特征优于文本特征的事件识别方法。
[Purpose/Significance]This paper proposes a method for network public opinion event recognition of sudden natural disasters based on syntactic features.[Methods/Process]The training set was constructed through data collection and event semantic annotation,and then a syntactic feature extraction method for network public opinion event recognition of sudden natural disasters was proposed.The syntactic feature extraction method was used to extract the event syntax from the training set,construct the event syntax feature library,and represent the event text in the form of syntactic vector;Finally,the syntactic similarity between event syntax and the syntax to be tested was used to calculate and identify events.[Results/Conclusion]Taking“typhoon lichima”as an example,through test set D_(2).It is determined that the optimal similarity of“typhoon lichma”event recognition was 0.93.Under this similarity,from the test set D_(2) 55 events and 82 non events were identified in test set D_(2)、D_(3) the F1 values of the experimental results were 0.851 and 0.929 respectively.At the same time,the comparative experiment shows that the syntactic feature is better than the text feature in the case of small training set.It provides a new reference for the research of network public opinion of sudden natural disasters.
作者
陈健瑶
夏立新
舒怡娴
Chen Jianyao;Xia Lixin;Shu Yixian(School of Information Management,Central China Normal University,Wuhan 430079,China)
出处
《现代情报》
CSSCI
2022年第6期17-26,93,共11页
Journal of Modern Information
基金
国家社会科学基金重大项目“新时代我国文献信息资源保障体系重构研究”(项目编号:19ZDA345)。
关键词
事件识别
突发自然灾害
网络舆情
句法相似度
event identification
sudden natural disaster
internet public opinion
syntactic similarity
