1.we use Tencent open source word vector：

URL：https://ai.tencent.com/ailab/nlp/data/Tencent_AILab_ChineseEmbedding.tar.gz

2.data:Chinese corpus

URL:https://github.com/thunlp/Chinese_Rumor_Dataset

3.Using attention mechanism and voting mechanism, the accuracy rate of the training set reaches 91.3%.