---
dataset_info:
  features:
  - name: streamer_id
    dtype: string
  - name: document_id
    dtype: int32
  - name: messages
    sequence: string
  - name: users
    sequence: string
  - name: timestamps
    sequence: float32
  - name: metadata
    struct:
    - name: user_view_count
      dtype: int32
    - name: user_broadcaster_type
      dtype: string
    - name: stream_game_id
      dtype: string
    - name: stream_type
      dtype: string
    - name: stream_viewer_count
      dtype: int32
    - name: stream_start_date
      dtype: string
    - name: stream_language
      dtype: string
  splits:
  - name: train
    num_bytes: 1197483835
    num_examples: 1951
  download_size: 510826866
  dataset_size: 1197483835
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
license: cc-by-4.0
---

<p align="center">
  📊 <a href="https://osf.io/bys56">Dataset</a> | 📄 <a href="https://osf.io/fqrj9">Paper</a>
</p>

# **TwitchChat Dataset**

**NOTICE** - Dataset originally published by Charlie Ringer, Mihalis A. Nicolaou, and James Alfred Walker on 2020-10-01, licensed under CC BY 4.0. This version is a non-modified port for distribution via Hugging Face Datasets. Original work available at [http://dx.doi.org/10.1609/aiide.v16i1.7439](http://dx.doi.org/10.1609/aiide.v16i1.7439).
This paper was published in Vol. 16 No. 1 (2020): Sixteenth AAAI Conference on Artificial Intelligence and Interactive Digital Entertainment.

The following information has been added from independent testing *without* the original authors. Inaccuracies may be present.

This dataset contains:
- A total of 666 unique streamers
- Approximately 14.7 million chat messages, which is approximately 162 million GPT-2 tokens.

The following information is recorded:
- Streamer IDs
- Document IDs
- Messages
- User IDs
- Timestamps
- Metadata


This dataset was ported to make it more accessible. No changes, moderation or filtering has taken place. The author on HuggingFace is not affiliated in any way with the original authors or original work. 
This dataset, and the accompanying paper were released under CC-By Attribution 4.0 International.

```
@article{
  Ringer_Nicolaou_Walker_2020,
  title={TwitchChat: A Dataset for Exploring Livestream Chat},
  volume={16},
  url={https://ojs.aaai.org/index.php/AIIDE/article/view/7439},
  DOI={10.1609/aiide.v16i1.7439},
  journal={Proceedings of the AAAI Conference on Artificial Intelligence and Interactive Digital Entertainment},
  author={Ringer, Charles and Nicolaou, Mihalis and Walker, James},
  year={2020},
  month={Oct.},
  pages={259-265}
}
```