baton4ik/danbooru-tag-metadata-snapshot

Danbooru Tag Metadata Snapshot

Public snapshot of Danbooru tag metadata for canonicalization, alias resolution, implication expansion, and analytics. It is available in both SQLite and Parquet for local indexed lookups, browser-based inspection, and general data tooling.

This dataset contains explicit Danbooru tag vocabulary, including sexual-content terminology, but does not include images or other media.

What Is Included

• canonical tags
• tag aliases
• tag implications
• snapshot metadata
• Parquet exports for viewer-friendly access

Files

• booru_snapshot.sqlite: ready-to-use SQLite snapshot
• tags.parquet: canonical tags table
• tag_aliases.parquet: alias mapping table
• tag_implications.parquet: implication mapping table
• metadata.json: human-readable metadata and row counts

Snapshot Metadata

• snapshot_built_at: 2026-04-08T09:15:30Z
• snapshot_builder_version: 0.4
• public_preparer_version: 0.5
• row_count_tags: 1711015
• row_count_tag_aliases: 59259
• row_count_tag_implications: 49334
• snapshot_profile: public_core

Sources

• live Danbooru tags API
• live Danbooru tag aliases API
• live Danbooru tag implications API

Schema

tags

• id: Danbooru tag id
• name: canonical tag name
• category: Danbooru numeric tag category
• post_count: current post count from the API snapshot
• is_deprecated: 1 if the canonical tag is deprecated, else 0
• words: serialized tokenized words array from the Danbooru API
• source: source label for this row in the snapshot build

Category values follow Danbooru's API:

• 0: general
• 1: artist
• 3: copyright
• 4: character
• 5: meta

tag_aliases

• id: Danbooru tag alias id
• antecedent_name: alias or redirected source tag
• consequent_name: canonical destination tag
• status: lifecycle state of the relation
• reason: optional human-entered explanation from Danbooru
• source: source label for this row in the snapshot build

tag_implications

• id: Danbooru tag implication id
• antecedent_name: source tag
• consequent_name: implied tag
• status: lifecycle state of the relation
• reason: optional human-entered explanation from Danbooru
• source: source label for this row in the snapshot build

Relation status values:

• active: current relation
• retired: historical relation retained for compatibility
• deleted: removed relation that should usually be ignored operationally

Practical Use

Use booru_snapshot.sqlite for local lookup-heavy applications, or use the Parquet tables directly in the Hugging Face viewer, DuckDB, Polars, Pandas, Spark, or other tabular tooling.

Example DuckDB query:

SELECT name, post_count
FROM read_parquet('tags.parquet')
WHERE category = 4
ORDER BY post_count DESC
LIMIT 20;

Example Python / Pandas:

import pandas as pd

tags = pd.read_parquet("tags.parquet")
aliases = pd.read_parquet("tag_aliases.parquet")