trentmkelly/archiveofourown-meta

Archive of Our Own (AO3) Fanfiction Metadata Archive Dataset

Introduction

This dataset contains metadata associated with approximately 13 million publicly available works hosted on Archive of Our Own (AO3). It has been compiled for research purposes, including natural language processing, data analysis, understanding trends within online creative communities, and the training of generative AI language models.

Crucially, this dataset contains only metadata and identifiers. It does not contain the full text or content of the referenced works from AO3.

To fetch the data yourself, you can take the id value from any row in the dataset and find the original work at https://archiveofourown.org/works/{ID}.

Data Format

The data is provided in JSONL format. Each line in the combined_metadata.jsonl file represents a single record and is a valid JSON object.

Each JSON object has the following top-level structure:

• id: (String) A unique identifier referencing the original work on AO3.
• title: (String) The title of the work.
• metadata: (Object) A nested JSON object containing various metadata fields extracted or associated with the work from AO3.

The metadata object contains key-value pairs describing the work. Common fields include, but are not limited to:

• author: (String) The author's listed name.
• Fandom/Fandoms: (String or Array of Strings) The source material fandom(s).
• Rating: (String) The content rating assigned.
• Category: (String) The primary relationship category (e.g., M/M, F/F, Gen).
• Relationship: (String or Array of Strings) Key relationship pairings.
• Characters: (String or Array of Strings) Key characters featured.
• Archive Warning: (String or Array of Strings) Content warnings provided by the author.
• Additional Tags: (String or Array of Strings) Freeform tags added by the author.
• Language: (String) The language of the work.
• published: (String) The publication date.
• completed: (String) The completion date, if applicable.
• chapters: (String) Chapter count (e.g., "1/1", "5/10").
• words: (String) The word count.

Legal Considerations and Copyright

This dataset consists solely of metadata and ID numbers that reference works publicly available on the internet via Archive of Our Own (AO3). No copyrighted creative works are stored or distributed in this dataset.

The collection and distribution of this metadata is based on established legal principles regarding hyperlinking and indexing under copyright law, particularly within the United States.

1. Linking vs. Copying: Providing a reference, identifier, or hyperlink to publicly accessible content on Archive of Our Own (AO3) is fundamentally different from reproducing or distributing the copyrighted content itself. US courts have generally recognized that linking does not constitute direct copyright infringement, as the entity providing the link does not host or transmit a copy of the work. This principle is exemplified by the "Server Test" articulated in cases like Perfect 10, Inc. v. Amazon.com, Inc., which distinguishes providing navigational instructions (like a link or ID) from the infringing act of hosting and displaying the content itself from one's own server. Distributing this dataset does not implicate the exclusive rights of reproduction or distribution under 17 U.S.C. § 106.

2. No Distribution of Works: This dataset contains factual metadata and identifiers associated with works on AO3, not the expressive content of the referenced works themselves. The legal analysis applicable to distributing copies of copyrighted material does not apply here.

3. Purpose: This dataset is intended as a resource for research, analysis, and potentially non-consumptive uses like training AI models on metadata patterns or language modeling based on publicly accessible data, consistent with principles of fair use where applicable.

Regarding Copyright Complaints:

This project operates under the well-established legal understanding that providing metadata and references (akin to links or citations) to publicly available online content, specifically works hosted on Archive of Our Own (AO3), does not constitute copyright infringement. The structure of this dataset—providing only metadata and identifiers, not the works themselves—is a deliberate choice to facilitate research while respecting copyright boundaries as understood under prevailing law.

Therefore, communications asserting copyright infringement based solely on the inclusion of metadata or an identifier referencing a work publicly hosted on Archive of Our Own (AO3) in this dataset will be considered legally unfounded. Such assertions demonstrate a misunderstanding of the distinction between linking/indexing and the distribution of copyrighted works. We will not engage in correspondence regarding claims based on the mere presence of metadata or an identifier for a work publicly hosted on Archive of Our Own (AO3) within this index.

Users who utilize the identifiers in this dataset to access the referenced works directly from Archive of Our Own (AO3) are solely responsible for ensuring their own subsequent use of those works complies with copyright law and the terms of service of Archive of Our Own (AO3).

Contact

For technical questions regarding the dataset format or structure, please open a discussion thread. Please review the legal considerations outlined above before initiating contact regarding copyright matters related to linking or indexing content from Archive of Our Own (AO3).