Dataset "anime_data.csv" is scraped from Bangumi anime site.
Those models apply Bert to predict the rating of new aniem based on title and summary in Bangumi anime site.
The most successful model is model4_upgrade one. The training result is in bottom of the python file.
Only support Chinese.