Research On The Concept And Relationship Extraction Of Public Opinion Ontology For Dynamic Topic

With the rapid development of the Internet age,the number of public opinion information is not poor,and the theme of new public opinion is constantly changing with time.How to grasp the public opinion information more quickly and accurately and analyze the tendency of public opinion are the main tasks of the government departments.The public opinion ontology can improve the semantic support of public opinion monitoring.At present,there are various forms of public opinion,such as news reports,micro-blog,post bar and so on.This thesis focuses on the study of public opinion ontology in news reports,which is not only standard in format,but also fast in timeliness.The existing domain ontology learning method based on public opinion research methods emerge in an endless stream,but the ontology learning has the characteristics of dynamic thematic and cross and other fields is not clear,this paper tries to find a suitable classification method to get the dynamic public opinion topic,then the public opinion and the relationship between the concept of ontology extraction based on the theme,in order to get the public opinion ontology learning method the support is beneficial for the later public opinion monitoring.The existing domain ontology learning method based on public opinion research methods emerge in an endless stream,but the ontology learning has the characteristics of dynamic thematic and cross and other fields is not clear,this paper tries to find a suitable classification method to get the dynamic public opinion topic,then the public opinion and the relationship between the concept of ontology extraction based on the theme,so as to obtain the public opinion ontology learning method the reference value for later public opinion monitoring.The main work of this paper is as follows:(1)according to the existing dynamic topic discovery stage,on the basis of the adjacent window to extract topic words words growth,ignore the integrity problem,proposed method combined with the precursor period and the subsequent public opinion words change trend and periodic heat topic extraction.(2)on the basis of improving the topic relevance of the thematic word clustering,candidate theme concept,in order to meet the theme of public opinion,public opinion by the definition of the concept of ontology theme membership filtering concept;then the concept for the seed word for compound words in the form of public opinion the concept of ontology.(3)the concept of relation extraction,combined with the concept of the distance between words and pattern matching method to classify relation judgment,pattern matching method can judge the word distance method can't determine the concept of;for the rest of the concept of extended sentence window,extracting sentence window conforming to the requirements of the verb,verb semantic filter selected.As the concept of the given semantic labels are non taxonomic relations,have been identified for the non linguistic rules classification relationship using extended to gain some unable to judge the relationship of.(4)the experiment proved that this method can effectively divide the public opinion topic and topic related public opinion extraction and ontology concept,the relationship between the concept of judgment,to provide reference for the construction of automatic and semi automatic construction of ontology of public opinion.