Research And Implementation Of Text Categorization Of Network Public Opinion Based On Markov Logic Networks

Along With the population and application of computer science and internet technology,the popularity of mobile devices,information is spreadingat an unprecedented rate.Nowadays people at all levels of society has urgent needs in identifying the content of public opinion.Beyond the traditional classification methods,since the Markov logic network proposed,it has been great concern of scholars.By studying the Markov logic network,data mining,machinery learn,human language process together with other aspects has got achievement frequently.In the existing network of public opinion research,we mainly face two questions as following:(1)All the time there are huge amounts of data on the Internet,it is impossible for all the data in a timely manner are marked with labels.AndClassification Code is based on a characteristic of the classification label scene.(2)If using the crawler,after screening and filtration,the limited data can really be intercepted.However Information missing problems has a great impact on the accuracy of the classification.For the current bottleneck stage of Classification problem,the thesis studies from the principle of Markov logic network,proposed a new classification methods based on Markov logic network using knowledge got from multiple sources and applying in target domain.And based on this technology,a prototype system has been designed to verify the classification results.Multi-source domain knowledge can be used to extract frequent common knowledge structure and apply in the target domain information missing conditions.This method overcome the lack of information.And the method use source domain knowledge to classified the target domain text context.By analysis the test results of our system it has been proved that our approach has certain advantages compared with single-source domain knowledge migration algorithm and without knowledge migration algorithm.Finally,we talk about some follow-up research of the system in the future to make the system efficient and accurate.