The purpose of tags is to help users find works. The purpose of tag wrangling is to group tags in the manner that most effectively enables users to find what they are looking for. The guidelines are designed to help Tag Wranglers determine what that manner should be.
When applying the tag wrangling guidelines to the actual tags we find on the archive, we run into a lot of edge cases: things that aren’t quite covered by the guidelines, or which for one reason or another the guidelines don’t seem like the best solution. In such cases, it can be helpful to refer back to the reasoning that underlines the guidelines themselves, for direction. The principles outlined here are the base ideas that inform the tag wrangling guidelines. However, sometimes they may conflict, and in reality the guidelines are a balancing act between these principles. In wrangling, always defer to the specific guidelines, where possible, rather than these more general principles. When developing new guidelines, consider how these principles intersect.
These Principles are descriptive, not prescriptive, and they are subject to a scheduled review by Tag Wrangling Supervisors to ensure that they continue to reflect current wrangling approaches.
These principles represent the wrangling philosophy and should guide how wranglers think about their work. They do not apply as written to individual tags, but are meant to apply to guideline change discussions and development.
In all things, we are guided by what the users do, not what we might wish they would do. We do not pass judgments on tags that do not violate the ToS, and we do not alter tags except where necessary for the technical operation of the site or to change capitalization or diacritics to match current guidelines for canonization. We use the Rule of Three not just to gauge what tags are too low-use to benefit from canonization, but also to remind us that if plenty of users are tagging for a concept we should consider whether this indicates that the concept is of use for filtering in a way not immediately apparent. Likewise, we respect the distinctions users are making with their tags. We don’t syn the specific to the general, because that erases nuances the users intended with their tags.
It is helpful to everyone if the formatting of canonical tags is consistent and predictable. When making new canonical tags, try to format them along the same pattern as other, similar canonicals.
No two tags can be identical, so if a given (canonical) tag is being used to refer to two completely unrelated concepts, disambiguated versions should be created if the different concepts can be clearly articulated. As much as possible, ambiguous tags should not be canonized.
We canonize tags because they are useful for filtering. We do not create new canonicals or tag trees only to organize existing tags, and not all related concepts must be linked if users are not tagging for them in that way. Tags should be canonized in a clearly recognizable form. When considering metatagging or synning a tag, ask yourself, “Would people looking for thing A always also want results for thing B? Are there circumstances where people looking for thing B would not want results for thing A as well?”
Avoid concept drift: Concept drift occurs when tags are added to a lower level of a tree that do not relate to the top layer. Always look at the full metatag tree before subtagging tags to avoid conceptual drift for tags higher up.
Respect for Technical Limitations:
We understand that the Ao3 Servers have previously not always been capable of supporting the user load, and we continue to design our guidelines to minimize strain imposed by tagging structures. We must also keep the archive's code structure in mind when choosing formats for canonical tags, and where possible avoid canonizing tags that won't work in some parts of the archive.