Paragraph Vector Model And Its Application Based On Penalty Function

The emotional analysis of microblogging data has important application value.We can extract public's concerning hot topics and their emotional tendency in these topics.But the microblogging data always have the problems of sparsity and high noise,how to structure the text data has been a matter of public.By variable selection method and the theories of social sciences,this article introduces the thought of coefficience shrinkage and social networks information to paragraph vector model,and construct the improved paragraph vector containing the semantic and emotional information.By the way,we test the model's validity by three English Twitter datasets.From the results,we can find that our model is better than other three traditional model.Finally,we crawled a lot of hot topics data from Sina Weibo,and structured the text data by our paragraph vector model.Simultaneously,we fit a text cluster model to extract the core content of hot topics as soon as possible.