douyin / README.md
tyhx's picture
Duplicate from bendavidsteel/douyin
05d10a2
|
Raw
History Blame Contribute Delete
3.42 kB
metadata
license: cc-by-nc-4.0
task_categories:
  - text-classification
  - feature-extraction
language:
  - zh
size_categories:
  - 1M<n<10M

Douyin Posts Dataset

A dataset of 1,133,545 unique posts from Douyin (Chinese TikTok), collected via snowball sampling of related videos.

Dataset Description

This dataset contains metadata from Douyin short videos, collected by crawling related video recommendations. Starting from seed videos matching specific keywords, the crawler iteratively fetched related videos to build a diverse sample of the platform's content.

Collection Method

  • Snowball sampling: Starting from keyword-matched seed videos, related videos were iteratively crawled
  • Deduplication: All posts are deduplicated by aweme_id (unique video identifier)
  • Partitioning: Data is partitioned by the first 4 digits of aweme_id for efficient access

Dataset Structure

The dataset contains 187 columns. Key fields include:

Core Fields

Field Type Description
aweme_id string Unique video identifier
desc string Video description/caption
caption string Short caption
create_time int64 Unix timestamp of video creation
duration int64 Video duration in milliseconds
region string Geographic region code

Engagement Statistics (statistics struct)

Field Description
digg_count Number of likes
comment_count Number of comments
share_count Number of shares
play_count Number of views
collect_count Number of saves/bookmarks

Author Information (author struct)

Field Description
uid Author user ID
nickname Display name
sec_uid Secondary user ID
signature Author bio
follower_status Follower relationship status

Content Metadata

Field Type Description
cha_list list Hashtag/challenge information
text_extra list Extracted hashtags, mentions, and links
music struct Audio/music metadata
video struct Video file metadata (URLs, dimensions, formats)
poi_info struct Location/point of interest data

Usage

import polars as pl
from huggingface_hub import hf_hub_download
from pathlib import Path

# Download a single partition
file_path = hf_hub_download(
    repo_id="bendavidsteel/douyin",
    filename="data/partition_7500.parquet.zstd",
    repo_type="dataset"
)
df = pl.read_parquet(file_path)

# Or load all partitions
from huggingface_hub import snapshot_download

local_dir = snapshot_download(
    repo_id="bendavidsteel/douyin",
    repo_type="dataset",
    allow_patterns="data/*.parquet.zstd"
)
df = pl.read_parquet(Path(local_dir) / "data" / "*.parquet.zstd")

File Structure

data/
├── partition_7035.parquet.zstd
├── partition_7040.parquet.zstd
├── ...
└── partition_7551.parquet.zstd

111 partition files, partitioned by the first 4 digits of aweme_id.

Limitations

  • Metadata only - does not include actual video/image content
  • Point-in-time snapshot - engagement statistics reflect collection time
  • Related video sampling may introduce biases toward popular/recommended content

License

CC-BY-NC-4.0 (Creative Commons Attribution-NonCommercial 4.0)