"Proposal and Evaluation of Japanese Toxicity Schema" provides a schema and dataset for toxicity in the Japanese language.
The repository is structured as follows:
schema.md: Labeling schema.data/subset.csv: A subset of the dataset used in "Proposal and Evaluation of Japanese Toxicity Schema". Annotation details are described inschema.md.
README_ja.md is written in Japanese.
data/subset.csv is composed of the following columns:
| Column Name | Description |
|---|---|
| id | Sentence ID |
| text | Sentence |
| Not Toxic | Toxicity Level: Not Toxic |
| Hard to Say | Toxicity Level: Hard to Say |
| Toxic | Toxicity Level: Toxic |
| Very Toxic | Toxicity Level: Very Toxic |
| category_卑語 | Toxicity Category: Dignity |
| category_差別 | Toxicity Category: Discrimination |
| category_迷惑行為 | Toxicity Category: Harassment |
| category_猥褻 | Toxicity Category: Obscenity |
| category_出会い・プライバシー侵害 | Toxicity Category: Privacy |
| category_違法行為 | Toxicity Category: Illegal |
| category_偏向表現 | Toxicity Category: Bias |
| annotation_num | Number of Annotators |