Research On The Training Of Automatic Chinese Hot Recommend Robot

In recent years,with the rapid development and popularity of social network and Mobile Internet,microblog has been widely used,typified by Twitter and Sina Weibo.Users can publish information(photos、videos、messages),share resources and discuss with each other on some topics anytime、anywhere.For this reason,microblog has been producing huge volumes of data containing users’ opinion.In a sense microblog has become highly lethal opinion carrier.The users tend to be more concerned about current hot topics and the real-time progress of hot evens.So it’s a mainstream research that how to use these data to timely discover hot topic event,track hot topics event in real time.Using the techniques of topics detection and tracing,we have designed and implemented a system of training of automatic Chinese hot recommend robots.The main researches are as follows:(1)The material collection module,which extracts microblog texts as material in real time using APIs of microblog open platform and then clean irrelevant or duplicate data,supplies enough raw material to the system.(2)The hot topics detection module,which detects the valuable information related to the topics on microblog constantly and accurately.The paper proposed a method of hot topic detection on microblog based on compatibility relation.The method that is quite different from traditional clustering algorithms and classification algorithms,uses mathematical model of compatibility relation to analyze the data of microblog.(3)The hot topics tracking module,which tracks the target information about specified topics from the subsequent data stream and shows the progress and evolution of topics.The paper proposed a self-adaptive microblog topic tracking method by user relationship.This method can closely track the development of topic and the evolution of focus with a high stability.(4)Automatic posting module,which aim at training the automatic Chinese hot recommend robot to publish the material related to the hot topic.Using APIs of microblog open platform,this paper has implemented a system of training of automatic Chinese hot recommend robots.The experiments prove that the material collection module,the hot topics identification module,the hot topics tracking module and the automatic posting module can work effectively.On the one hand,by dececting and tracking the hot topic from complex data stream of microblog,the system provides the precondition to guid and control the public opinion of microblog.On the other hand,by automaticly publishing the reports related to the topic,the system make users conveniently grasp the overall trend of topic and achieves the recommendation based on the user’s interest.