Bilibili weekly picks dataset links videos with comments and covers
BiliPicks: Seven Years of Weekly Platform Curation on Bilibili with Video Metadata, Comments, and Danmaku
Social and Information Networks
Summary
It can be hard to understand how video platforms choose and promote videos and how audiences respond over time. The authors collected seven years of weekly video selections from Bilibili, including detailed video data, millions of comments, time-synced chat comments called danmaku, and cover images. This large dataset helps show patterns in video themes, creators, and how viewers react to specific moments. It provides a way to study how platform editors shape viewing trends and discussions.
What this means in practice
- •For social media analysts: Analyze long-term changes in video selections and audience responses on Bilibili using linked metadata and comments.
- •For video content creators: Identify trends in popular themes and audience engagement through curated weekly picks to guide content strategy.
Authors
Ruomeng Liu
Abstract
BiliPicks documents seven years of Bilibili's Weekly Picks, a platform-curated editorial video series. The dataset covers 371 publicly listed issues and 13,758 curated videos from 22 March 2019 to 7 May 2026. It links issue dates, themes and editorial recommendations to video metadata, 47,799,657 top-level comments, 125,690,760 time-synchronised overlay comments (danmaku), and 13,739 cover images. Weekly issues provide a consistent unit for tracing changes in category composition, recurring creators and editorial framing. Corresponding comments and danmaku connect video-level discussion with audience expressions tied to particular moments, while covers add a visual dimension. We evaluate retrieval coverage and illustrate longitudinal analyses of selection, audience language and cover presentation. BiliPicks provides a dataset for future studies of platform curation, linking changes in published selections to their editorial framing and corresponding audience engagement.