Relevance Identification of Chinese News in the New Media Environment - Taking 'Shandong Vaccine Event' as an Example

摘要

We construct the maximum entropy model, which improved the feature function and training set, to calculate the relevance degree of Chinese news and selected topics, and apply them to Chinese news event relevance recognition. According to the example of "Shandong vaccine event", the relevant data are obtained on the micro-blog platform. By selecting different numbers of features, this paper compare the maximum entropy model with the support vector machine, BP neural network, Bayes and K-means algorithm, which are four kinds of common text classification method of micro-average accuracy, by empirical analyzing. Experiments have found that although the method is not always superior to the support vector machine method, it is superior to the other three methods.