This repository contains the character type identification dataset.

For more details, see the paper TBD.

• documents.csv - contains document metadata from document_id, set, script_url, script_file_size, script_word_count, script_start, script_end, wiki_url, wiki_title.
• summaries.csv - contains wikipedia summaries in the format document_id, set, summary.
• character_labels.csv - contains the character type annotations in the format document_id, set, character_name, character_type
• download_scripts.py - downloads the full scripts.

Due to licensing issues, the full scripts aren't included in this repository. They can be downloaded to /path/to/repo/tmp by running:

pip install -r requirements.txt
python download_scripts.py

Alternatively, the dataset can be conveniently loaded using huggingface/datasets:

import datasets
ds = datasets.load_dataset("character_type_id")
print(ds["train"][0])
>> {"document_id": "00001", "summary":{"title": "Name of Movie (film)", "text": "The movie begins..."},...

@article{characterTypeID,
author = {TBD},
title = {TBD},
journal = {TBD},
url = {https://TBD},
year = {2021},
pages = {TBD},
}