ao3 — Archive of Our Own fan fiction works and tags (2021)

Description

A bipartite network of the works of fan fiction published on the Archive of Our Own (AO3), the fan works repository of the Organization for Transformative Works (OTW), and the tags attached to them, from the one-time data dump released by the OTW on 26 February 2021 "for fan statisticians". It covers all the works on AO3 at the time, from 2008 to February 2021, except those hidden by the site (e.g. drafts and works hidden by its administrators).

An edge connects a work to each of its tags. AO3 tags describe the works along several dimensions, each a tag type: the rating (e.g. General Audiences, Explicit), the archive warnings, the categories of relationships (e.g. F/M, M/M, Gen), the fandoms (the source works, e.g. Harry Potter - J. K. Rowling), the characters, the relationships between characters (e.g. Draco Malfoy/Harry Potter), and free-form tags (e.g. Fluff, Alternate Universe). Tags are written freely by the authors, and the volunteer "tag wranglers" of AO3 mark one tag of each set of synonyms as canonical, and the others as synonyms of it (merged into it); the works are linked to the tags as written by their authors, and the canonical tag of a tag is given by the property merger_id.

To protect the privacy of the authors, the names of the tags used fewer than five times were redacted by the OTW (their name is Redacted, about 12.8 million of the 14.5 million tags). The works carry no identifiers, titles or authors, and are numbered in the order of the dump (roughly by decreasing creation date). Only the tags used by some work are nodes of the network (tags that are only canonical versions of other tags, without works of their own, are left out; they are still identified by merger_id). 8,139 tags used by works are missing from the tag file of the dump; their name and type are empty, canonical is false, and cached_count and merger_id are -1. Three empty items in the tag lists of the works are left out.

The network has the following properties:

(Node) is_work: whether the node is a work (otherwise it is a tag).
(Node) time (works): the creation date of the work, as Unix time.
(Node) language (works): the language of the work, as a code (e.g. en, zh).
(Node) restricted (works): whether the work is visible only to registered users.
(Node) complete (works): whether the work is complete.
(Node) word_count (works): the number of words of the work (-1 if not given).
(Node) tag_id (tags): the identifier of the tag on AO3 (-1 for works).
(Node) name (tags): the name of the tag (Redacted if it was used fewer than five times).
(Node) type (tags): the type of the tag: Rating, ArchiveWarning, Category, Fandom, Character, Relationship, Freeform, Media or UnsortedTag (tags not yet sorted by the wranglers).
(Node) canonical (tags): whether the tag is canonical.
(Node) cached_count (tags): the approximate number of uses of the tag, as given by AO3.
(Node) merger_id (tags): the identifier of the canonical tag of which the tag is a synonym (-1 if none).

Tags

Social Online Folksonomy Bipartite Unweighted Metadata

Added

2026-10-28

Citation

Organization for Transformative Works, "Selective data dump for fan statisticians." Archive of Our Own (2021) archiveofourown.org

Upstream URL 418

https://archiveofourown.org/admin_posts/18804

Networks

Tip: hover over a column header for a description.

Column legend

$\left<k\right>$: Mean degree
$\sigma_k$: Standard deviation of the degrees
$\lambda_h$: Largest eigenvalue of the Hashimoto (non-backtracking) matrix
$\tau$: Random walk mixing time
$r$: Degree assortativity
$c$: Global clustering coefficient
$\oslash$: (Pseudo-) diameter
$S$: Relative size of the largest component
NPs: Node properties
EPs: Edge properties

Name	Nodes	Edges	$\left<k\right>$	$\sigma_k$	$\lambda_h$	$\tau$	$r$	$c$	$\oslash$	$S$	Kind	Mode	NPs	EPs	gt	GraphML	GML	csv
ao3	18,484,023	119,146,249	12.89	1768.82	2423.45	54.99	-0.09	0.00	6	1.00	Undirected	Bipartite	is_work time language restricted complete word_count tag_id name type canonical cached_count merger_id		260.3 MiB	687.0 MiB	628.0 MiB	623.5 MiB