文本主题可视化及其在上市公司风险分析中的应用

• 作 者: 赵一鸣 张进
• 机 构: 武汉大学信息管理学院,美国威斯康辛大学密尔沃基分校信息研究学院,
• 出 处: 图书情报工作
• 发表日期: 2014年02期
• 关键词: 可视化 知识发现 主题发现 主题可视化 文本主题 邻近关系 空间图 上市公司 风险识别 词汇

使用可视化的方法,从词汇的粒度去发现和描述主题,指出属于同一个主题的词汇之间存在集聚特性,使用转置向量空间中的邻近关系表示这种集聚特性,进而使用多维尺度分析的方法将词汇之间的邻近关系投影到低维的MDS空间图中,以使有集聚关系、从属于同一个主题的词汇在空间图中聚类成主题。最后,将该方法成功应用于计算机应用服务业上市公司的风险分析,揭示该行业市场风险的具体表现和语义内容。

Using visualization method, the topics are found and described from the granularity of the lexicon. It is pointed out that there are some agglomeration features among the words belonging to the same subject, and the adjacent features in the transposed vector space are used to represent the agglomeration characteristics. Then the method of multidimensional scaling analysis The adjacent relations between words are projected into the low-dimensional MDS space map so that there is a clustering relationship, and terms belonging to the same topic are clustered into topics in the space map. Finally, the method is successfully applied to the risk analysis of the listed companies in the computer application service industry, revealing the specific performance and semantic content of the market risk in the industry.