基于动态主题模型的舆情本体概念抽取

Concept extraction of public opinion ontology based on dynamic topic model

对现有领域本体学习方法不能直接移植在舆情本体学习方法上的问题进行研究,根据舆情本体的动态主题性提出一种基于动态主题模型的舆情本体概念抽取的方法。结合舆情周期内词语的前驱增速和后继增速提取主题特征词,主题度选取主题词;对主题词通过主题相关度进行聚类,形成主题词簇;对主题词簇下候选概念进行主题概念隶属度的过滤得到舆情本体概念;以上述概念作为种子词寻找左右邻接词得到复合词,过滤得到舆情本体复合概念。实验结果验证了该方法具有更高的准确率和召回率。

The existing domain ontology learning method can not be directly transplanted into the study of the sensation ontology learning method.After studying the problem,a method of extracting the concept of public opinion ontology based on the dynamic topic model was proposed according to the dynamic theme of the public opinion ontology.Thematic feature words were extracted based on the predecessor growth rate and subsequent growth rate of the words in the lyric cycle,and the topic degree was used to select the topic words.The topic words were clustered by topic relevance to form the topic word clusters.The subject concept word cluster was filtered under the theme concept membership degree to obtain the concept of public opinion ontology.The concept above was used as the seed word to find the left and right adjoining words to get compound words,and the concept of lyric ontologies was filtered.Experiments verify that the proposed method has higher accuracy and recall rate.