Research And Implementation Of Weibo Rumor Detection Based On Comment Attention Mechanism

Due to the rapid development of mobile Internet and social media,people have more and more means to obtain information.According to the different recognition of information sources,the credibility of information is also questionable.At present,while mobile social media can quickly and widely spread messages,there are many people who deliberately spread malicious false messages that cause damage to society by taking advantage of the convenience of Internet information transmission.Considering that the traditional rumor detection model based on deep learning is usually based on manual construction of text features,which is subjective.However,the model data set based on the neural network has limited topics and lacks certain generalization and interpretability.To solve the above problems,this paper will design an interpretable rumor detection model that automatically constructs features and enhances the semantic representation of knowledge:1.An Ernie(enhanced representation through knowledge integration)pre-training model based on enhanced knowledge learning is proposed.The model can learn the context semantic relationship more accurately by encoding the entity and entity-relationship knowledge maps respectively;2.A multi-level comment blog common attention mechanism network is constructed to learn from the words and sentences of blog posts and comments.Finally,the classification results and the most weighted words and phrases in the comments with the strongest correlation with the original blog are obtained as the explanatory evidence of the model;3.The prototype system integrating the above rumor detection model is designed and implemented,and its high concurrency and high availability are verified in the final test.This model is used in CED_Dataset,and the accuracy rate on the dataset is 89.40%,the F1 score is 0.8837,and the recall rate is 94.37%,which is higher than that of other models in the same dataset.