lukeslp/bluesky-alt-text-observatory

Bluesky Accessibility Observatory

This is a focused longitudinal observation of declared image descriptions in public Bluesky post commits. It begins with archive-format v2 and does not include the biased April 2026 snapshot corpus.

daily_metrics and daily_language_metrics are aggregate observations at post creation time. description_sample is a deterministic, uniform bottom-k sample of non-empty descriptions after a 48-hour correction window. It uses keyed pseudonyms, not anonymous identifiers. Deletes and inactive-account markers remove matching sample rows when observed; aggregate historical counts remain creation-time observations.

Jetstream v1 is at-least-once and lacks sync markers, so this is responsive to post/account deletion markers, not a perfect mirror of present network state. Whitespace-only alt is missing; a multi-image post is fully described only when every image has non-empty alt. Languages are declared primary BCP-47 tags or unknown; no inferred language or English-only filtering is used.

Aggregate metrics and schema may be reused under CC0. Sampled descriptions retain their authors' rights; this dataset does not claim a CC-BY license for Bluesky user text. Do not use pseudonyms to profile people or join the data to other identity-bearing sources.

Sampling weight is eligible non-empty descriptions that UTC day / published descriptions that UTC day; at most 2,000 rows per day are published. See docs/METHODOLOGY.md, docs/DATA_DICTIONARY.md, docs/CHANGELOG.md, and manifest.json for coverage, fields, freshness, checksums, and loading examples.