Research on deep neural network based classification method for social robots

we propose a classification method for social bots in this paper. The microblog comment text topic is first extended so topic can express the event content more effectively. Then a generative adversarial network model is used to determine the relevance of the microblog text and the extended topic to detect contaminants. Then the social bot opinion sentence generation rule is combined with the deep learning model TextCNN to get the opinion sentence recognition method to further distinguish the commenters from the spreaders. Finally, migration learning is used to train the model to improve the classification effect.