honeray/ai-music-comments-1.5M

---
license: afl-3.0
language:
- en
tags:
- ai music
- sentiment analysis
- asba
- comments
task_categories:
- text-classification
- feature-extraction
pretty_name: 1.5M Comments on AI Music
size_categories:
- 1M<n<10M
---
## What this is
A corpus of >1.5M comments about AI music gathered from YouTube, Bilibili, Reddit, Twitter/X, and Weibo

## What the files are
`raw_data`: the full dataset of >1.5M comments

`proc_data`: the processed dataset (around 221K comments) that has filtered out any comments that either
- have <5 likes, or
- are under a video post (i.e. YouTube or Bilibili) with a views to comments ratio less than 0.001

`final_data`: the final dataset (1363 comments) with both a comment relevance and post relevance score >= 0.6

## What the features are

`post_title`: the title of the post

`post_views`: the number of views that the post has

`post_likes`: the number of likes that the post has

`creator_id`: the ID of the creator of the post

`creator_subs`: the number of subscribers the creator of the post has (in the case of YouTube or Bilibili)

`post_comments`: the total number of comments under the post that the current comment was published under
- this includes 弹幕 (bullet comments) in the case of Bilibili

`comment`: the actual content of the comment itself

`comment_likes`: the number of likes that the individual comment received (may be negative in the case of Reddit)

`platform`: the social media platform where the comment was published

`url`: the url of the post where the comment was published

`comment_id`: the ID of the comment

`retweets`: the number of retweets that the comment received (in the case of Twitter/X or Weibo)
- In the case of Weibo, the number of retweets is no longer an exact amount after 1 million; for posts with more than 1 million retweets, 1 million is used instead.

`replies`: the number of replies that the comment received

`quotes`: the number of times that the comment was quoted

`bookmarks`: the number of times that the comment was bookmarked

`is_retweet`: whether the comment is a retweet of some original tweet

`is_quote`: whether the comment is a quote from some original tweet

`comment_parent_id`: the ID of the comment that this comment replied to

`post_desc`: the description (if any) of the post that this comment was published under

`post_dislikes`: the number of dislikes that the post the comment was published under received

`post_favs`: the number of times that the post the comment was published under was favorited
- this includes coins in the case of Bilibili

`post_shares`: the number of times that the post the comment was published under was shared

`post_keyword`: the keywords of the post the comment was published under

`creator_likes`: the number of likes that the comment received from the creator of the post

## More Information
For more information on how these comments were collected, please see the associated paper at (link coming soon)

---
license: cc-by-nc-4.0
---