Font Size: a A A

Analysis And Presentation Of News Hotspots For Multimedia Questions And Answers

Posted on:2018-03-13Degree:MasterType:Thesis
Country:ChinaCandidate:T CaoFull Text:PDF
GTID:2358330512976689Subject:Computer technology
Abstract/Summary:PDF Full Text Request
In recent decades,with the popularity and development of Internet technology,a large number of news reports emerged on the Internet.In the face of a large number of news and information,users need to take a long time to understand the news hot topic,so how to help users quickly understand the news Topic has become an urgent need to solve the problem.In this paper,we propose a hot topic detection method based on semantic feature and a topic evolution method based on key documents and LDA model,which includes the following points:1)A hot news detection method based on semantic feature is proposed.This method takes into account the different effects of news headlines and news texts on the results of news topic.detection,and uses subject model to extract the semantic features of the title and the text of news respectively.The semantic features include document-theme matrix and theme-word matrix.Then,we use VSM to represent the document and adjust the feature weights of VSM model according to the semantic feature.Then we use the improved clustering algorithm proposed in this paper to cluster and express the topic through clustering center.Finally,we use the news Heat calculation formula to calculate the heat of hot during the specified time,and then we use the calculated value of the heat to sort hot news list in during the specified time.2)A topic evolution method based on key documents and LDA model is proposed.Using the LDA model for the document set,the global topic-document and subject-word probability distributions are obtained from the document set.The LDA model is then used to extract the document-topic and topic-word probability distributions in each time window,and each window is extracted according to the key document definitions The key documents of each window are represented by the global document-topic probability distribution,and then the similarity of the key documents of adjacent windows is calculated.By the similarity of the key documents,the same topic of different windows is related and finally the evolution process of the topic is obtained.3)Realize the Internet news hot topic detection and tracking system.The system following the MVC design pattern,mainly achieves the news hot spots detection and news hot topic evolution function.The various functional modules are independent of each other to update and expand the system.
Keywords/Search Tags:Semantic features, Topic detection and tracking, Topic evolutionary algorithm, News heat calculation
PDF Full Text Request
Related items