Traditional rumor detection methods rely on artificial features, which is inefficient and weak in generalization. Recurrent neural network has obvious advantages in processing sequential data, but gradient disappearance is difficult to solve. Aiming at the above problems, this paper proposes a microblog rumor detection method based on Transformer model. This method adopts the word embedding method of XLNet, extracts deep semantic features from microblog books through the encoder of Transformer, and then inputs the learned deep semantic features into Softmax layer to get the final classification result, and then realizes microblog rumor detection.
展开▼
