Datasets:

TrustAIRLab
/

OtomeSCAN

pretty_name: OtomeSCAN
license: cc-by-nc-4.0
language:
  - zh
  - en
task_categories:
  - text-classification
size_categories:
  - 1K<n<10K
extra_gated_heading: Researcher Access Agreement
extra_gated_description: Access requests are typically reviewed within 1–3 days.
extra_gated_button_content: Acknowledge and request access
extra_gated_prompt: |
  This dataset contains social media posts from otome game communities,
  including potentially offensive content. By requesting access, you agree
  to use the dataset solely for academic purposes and not to attempt to
  re-identify users represented in the dataset.
extra_gated_fields:
  Full Name: text
  Organization: text
  Organization Email: text
  Intended Use:
    type: select
    options:
      - Research
      - Education
      - label: Other
        value: other
  I will not attempt to re-identify users represented in the dataset: checkbox
tags:
  - toxicity
  - otome-games
configs:
  - config_name: weibo
    data_files: weibo.csv
  - config_name: reddit
    data_files: reddit.csv

OtomeSCAN

Human-annotated posts from otome-game communities, labelled for toxicity and for who the toxicity targets. From the paper “Shut Up and Let Me Enjoy My Otome”: Understanding and Measuring the Toxicity in Otome Game Communities, CCS 2026.

weibo.csv — Chinese posts from Weibo
reddit.csv — English posts from Reddit

Columns

weibo.csv

Column	Meaning
`community`	Community the post comes from: `LoveAndDeepspace`, `MrLove`, `TearsOfThemis`, `Genshin_Impact`, `otomegames`, `Games`
`content`	Full text of the post, verbatim
`Toxicity`	`0` = non-toxic, `1` = toxic
`Target_Type`	Who the toxicity targets, `A`–`H` (below); empty when `Toxicity = 0`

reddit.csv

Column	Meaning
`community`	Subreddit the post comes from, without the `r/` prefix; same six names as above
`title`	Submission title
`selftext`	Submission body; empty for link/image posts
`content`	`title` and `selftext` joined by a newline — the text that was annotated, and the counterpart of Weibo's `content`
`Toxicity`	`0` = non-toxic, `1` = toxic
`Target_Type`	Who the toxicity targets, `A`–`H` (below); empty when `Toxicity = 0`

Target_Type


`A` players	`B` NPCs	`C` game developers	`D` platform moderators
`E` policymakers	`F` identity groups	`G` other game-related entities	`H` unknown

Related release

Model-generated toxicity labels for all posts analysed in the paper are released on Zenodo: https://doi.org/10.5281/zenodo.22817364.

Citation

If you find this useful in your research, please consider citing:

@inproceedings{ZSJBZ26,
      author = {Yage Zhang and Xinyue Shen and Yukun Jiang and Michael Backes and Yang Zhang},
      title = {{``Shut Up and Let Me Enjoy My Otome'': Understanding and Measuring the Toxicity in Otome Game Communities}},
      booktitle = {{ACM SIGSAC Conference on Computer and Communications Security (CCS)}},
      publisher = {ACM},
      year = {2026}
}