基于短文本的舆情本体概念的抽取研究

Research on Public opinion Ontology Concept Extraction Based on short text

针对现有的本体学习方法难以适应短文本类型新闻媒体资讯由原来的长文本转为短文本的情形。本文提出了一种"基于短文本的舆情本体概念的抽取方法"。短文本语料经过词频统计,其词频呈现数据稀疏现象,经过词频调整后参加后续概念抽取过程。短文本的描述信息较差。如果采用向量空间模型来表征短文本会造成向量空间的语义缺失和高维稀疏,所以本文使用集合空间来计算文档相似度,抽取与主题相关的核心概念。实验结果理想。

In view of the existing ontology learning method, it is difficult to adapt to short text type news media information from orig-inal long text to short text. This paper proposes a "short text based ontology concept extraction method". After word frequency statis-tics, short word frequency data show sparse data. After word frequency adjustment, it takes part in subsequent concept extractionprocess. The description information of short text is poor. If using vector space model to characterize short text will cause semanticloss and high dimension sparsity in vector space, this paper uses set space to calculate document similarity and extract the coreconcepts related to the subject. The results of the experiment are ideal.