VTuber 1B: Live Chat and Moderation Events

VTuber 1B is an academic purpose NLP dataset, collecting over a billion live chats, superchats, and moderation events (bans/deletions) from virtual YouTubers' live streams.

Download the dataset from Kaggle Datasets and join #livechat-dataset channel on holodata Discord for discussions.

We also offer ❤️‍🩹 Sensai, a live chat dataset specifically made for building ML models for spam detection / toxic chat classification.

Provenance

• Source: YouTube live chat events collected by our Honeybee cluster. Holodex is a stream index provider for Honeybee which covers Hololive, Nijisanji, 774inc, etc.
• Temporal Coverage:
  • Chats: from 2021-01-15
  • Super chats: from 2021-03-16
  • Super stickers: from 2022-01-20 (N/A yet)
  • Membership joining events: from 2021-10-18 (N/A yet)
  • Membership milestones: from 2021-10-20 (N/A yet)
  • Membership gifts: N/A
  • Placeholders: from 2022-01-21 (N/A yet)
• Update Frequency:
  • At least once every 6 months

Research Ideas

• Toxic Chat Classification
• Spam Detection
• Demographic Visualization
• Superchat Analysis
• Training neural language models

See public notebooks built on VTuber 1B and VTuber 1B Elements for ideas.

We employed Honeybee cluster to collect real-time live chat events across major Vtubers' live streams. All sensitive data such as author name or author profile image are omitted from the dataset, and author channel id is anonymized by SHA-1 hashing algorithm with a grain of salt.

Editions

VTuber 1B Elements

Kaggle Datasets (2 MB)

VTuber 1B Elements is most suitable for statistical visualizations and exploratory data analysis.

VTuber 1B

Kaggle Datasets (47 GB)

VTuber 1B is most suitable for frequency analysis. This edition includes only the essential columns in order to reduce dataset size and make it faster from Kaggle Kernels to load data in.

VTuber 1B Complete

VTuber 1B Complete is only available to those approved by the admins. If you are interested in conducting research using this edition, please reach us at [email protected] (for organizations only).

Dataset Breakdown

Ban and deletion are equivalent to markChatItemsByAuthorAsDeletedAction and markChatItemAsDeletedAction respectively.

Channels (channels.csv)

Inactive channels have INACTIVE in group column.

Pandas usage

import pandas as pd

dtype_dict = {
    'channelId': 'category',
    'name': 'category',
    'englishName': 'category',
    'affiliation': 'category',
    'subscriptionCount': 'int32',
    'videoCount': 'int16',
    'photo': 'category'
}
chats = pd.read_csv('../input/vtuber-livechat-elements/channels.csv', dtype=dtype_dict)

Chat Statistics (chat_stats.csv)

Pandas usage

import pandas as pd

chat_stats = pd.read_csv('../input/vtuber-livechat-elements/chat_stats.csv'))
sc_stats = pd.read_csv('../input/vtuber-livechat-elements/superchat_stats.csv'))
stats = pd.merge(chat_stats, sc_stats, on=['period', 'channelId'], how='left')

Super Chat Statistics (superchat_stats.csv)

Chats (chats_%Y-%m.parquet)

Membership status

Pandas usage

import pandas as pd

chats = pd.read_parquet('../input/vtuber-livechat/chats_2022-02.parquet')

Superchats (chats_:year:-:month:.parquet)

Color and Significance

Pandas usage

import pandas as pd
from glob import iglob

sc = pd.concat([
    pd.read_parquet(f)
    for f in iglob('../input/vtuber-livechat/superchats_*.parquet')
], ignore_index=False)
sc.sort_index(inplace=True)

Deletion Events (deletion_events.parquet)

Pandas usage

Insert deleted_by_mod column to chats DataFrame:

chats = pd.read_parquet('../input/vtuber-livechat/chats_2022-02.parquet')
delet = pd.read_parquet('../input/vtuber-livechat/deletion_events.parquet', columns=['id', 'retracted'])

delet = delet[delet['retracted'] == 0]

delet['deleted_by_mod'] = True
chats = pd.merge(chats, delet[['id', 'deleted_by_mod']], how='left')
chats['deleted_by_mod'].fillna(False, inplace=True)

Ban Events (ban_events.parquet)

Here Ban means either to place user in time out or to permanently hide the user's comments on the channel's current and future live streams. This mixup is due to the fact that these actions are indistinguishable from others with the extracted data from markChatItemsByAuthorAsDeletedAction event.

Pandas usage

Insert banned column to chats DataFrame:

chats = pd.read_parquet('../input/vtuber-livechat/chats_2022-02.parquet')
ban = pd.read_parquet('../input/vtuber-livechat/ban_events.parquet', columns=['authorChannelId', 'videoId'])

ban['banned'] = True
chats = pd.merge(chats, ban, on=['authorChannelId', 'videoId'], how='left')
chats['banned'].fillna(False, inplace=True)

Consideration

Anonymization

id and authorChannelId are anonymized by SHA-1 hashing algorithm with a pinch of undisclosed salt.

Handling Custom Emojis

All custom emojis are replaced with a Unicode replacement character � (U+FFFD).

Redundant Ban and Deletion Events

Bans and deletions from multiple moderators for the same person or chat will be logged separately. For simplicity, you can safely ignore all but the first line recorded in time order.

Citation

@misc{vtuber-livechat-dataset,
 author={Yasuaki Uechi},
 title={VTuber 1B: Large-scale Live Chat and Moderation Events Dataset},
 year={2022},
 month={2},
 version={37},
 url={https://holodata.org/vtuber-1b}
}

License

• Code: MIT License
• Dataset: ODC Public Domain Dedication and Licence (PDDL)