|
--- |
|
license: cc-by-4.0 |
|
--- |
|
# Ice and Fire Comment Dataset |
|
|
|
## Description |
|
|
|
The Ice and Fire Dataset is a collection of comments from the Icelandic blog platform, blog.is, that have been annotated in several tasks. |
|
|
|
## Dataset Structure |
|
|
|
### Data Fields |
|
|
|
- `annotator_id`: An integer identifier for the annotator who labeled the comment. |
|
- `label`: The label assigned to the comment. |
|
- `task_type`: The type of task the comment was annotated for (see paper). |
|
- `show_blog_post`: A boolean indicating whether the annotator viewed the blog post in the annotation process. |
|
- `show_preceding_comments`: A boolean indicating whether the annotator viewed preceding comments in the annotation process. |
|
- `blog_title`: The title of the blog post associated with the comment. |
|
- `blog_text`: The text of the blog post associated with the comment. |
|
- `comment_body`: The body of the comment. |
|
- `previous_comments`: A string containing all previous comments concatenated together, separated by " || ". |
|
|
|
### Data Splits |
|
|
|
This dataset is provided as a single CSV file, `ice_and_fire_huggingface_dataset.csv`, without predefined training, validation, or test splits due to the size and label distribution. Users are encouraged to create their own splits as needed for their specific tasks or to use cross-validation for benchmarking. |
|
|
|
### Citation Information |
|
|
|
If you use the Ice and Fire Dataset in your research, please cite it as follows: |
|
|
|
TODO |
