Weibo Fake News Dataset and Experiment Models

The rumdect folder contains the original Weibo fake news dataset (Weibo folder and Weibo. txt text) and the processed dataset used in this experiment (data, label0, and label1 folders). The original Weibo dataset is Jing Ma's Weibo dataset. This experiment uses a subset of the Weibo dataset, which includes 1280 real news and 1280 fake news, each with 80 comments. The saved-loss folder and saved-model folder are used to save the results and model parameters generated during the experiment process. Due to the large size of model parameters, they are not saved but can be obtained by running experiment code. The used_model folder includes the three models used in this experiment: chinese-bert-wwm, bert-base-personality, trans-opus-mt-zh-en. These three models can all be obtained from the hugging face website.code.zip includes all experiment codes.