Analysis Based On The Keywords And Network Topics Of The Time Evolution

With developing of network technology at full speed,the network medium has already been the main carrier by which people can gain all kinds of information,and has carried a large number of news reports.So we can find all kinds of topics reported on various websites.Normally,the reporting probability of hot topics is greater than of normal ones,and if the news repots about some topic are reported in different stages,the hanker degrees of different stages are much different.And by analyzing the differences we can know the hot topics in daily life and in some degree know the spirit and social life stat of people.The evolution analyzing of topics is an important branch in TDT and an important direction in network security technology.And by analyzing the evolution processing of some topic,we can know the main subtopics of the topic and some kinds of relations between the subtopics.And it is beneficial to build the models of topic evolution.For two types of document sets,we propose different methods to get the subtopics and analyze the evolution.And one is the result set returned by web search engines.And this kind of document is composed of a title and an abstract of web page behind the link.But the content of the abstract can't descript the topic and normally are the keyword sets.So we propose a method that is based on keywords to analyze the evolution of topics and draw the subtopics of some topic through the keywords gained by some method that can depict a set of documents.By analyzing the snapshots of document sets,we can get to know the relations between the subtopics and the evolution process of the topic.And the other document set is the news reports reported by Web Sizes,BBS,RSS and so on.And it has the entire story content and the time element is very easy to find,and there is an event following each time depiction showing.So we can make use of the time element to extract the subtopics and build the event model to express the topic and analyze the sub-vectors of the event vectors to know the evolution of the topic.And through the experiments,we find the two methods are well in drawing the subtopics and show the relations among the subtopics and the evolution correctly.But because of the differences between the characters of document sets,the method of drawing subtopics based on time point has a higher accurate rate than the one based on keywords and the model of the subtopics is more readable.