Replication Data for: Cross-Partisan Discussions on YouTube: Conservatives Talk to Liberals but Liberals Don't Talk to Conservatives

The dataset is first introduced in the following paper: Siqi Wu and Paul Resnick. Cross-Partisan Discussions on YouTube: Conservatives Talk to Liberals but Liberals Don't Talk to Conservatives. In AAAI International Conference on Weblogs and Social Media (ICWSM), 2021. us_partisan.csv Metadata for 1,267 US partisan media on YouTube. The first row is header. Fields include \"title, url, channel_title, channel_id, leaning, type, source, channel_description\" video_meta.csv Metadata for 274241 YouTube political videos from US partisan media. The first row is header. Fields include \"video_id, channel_id, media_leaning, media_type, num_view, num_comment, num_cmt_from_liberal, num_cmt_from_conservative, num_cmt_from_unknown\" user_comment_meta.csv.bz2 Metadata for 9,304,653 YouTube users who have commented on YouTube political videos. The first row is header. Fields include \"hashed_user_id, predicted_user_leaning, num_comment, num_cmt_on_left, num_cmt_on_right\" user_comment_trace.tsv.bz2 Comment trace for 9,304,653 YouTube users who have commented on YouTube political videos. The first row is header. Fields include \"hashed_user_id predicted_user_leaning comment_trace\" (split by \t) \"comment_trace\" consists of \"channel_id1,num_comment_on_this_channel1;channel_id2,num_comment_on_this_channel2;...\" (split by ;) trained_HAN_models.tar.bz2 Five trained HAN models for predicting user political leanings. Each model consists a \".h5\" model file and \".tokenizer\" tokenizer file. See this for how to use our pre-trained HAN models. See more details in this data description.