Seeking suggestions for tagging changes.

Seeking suggestions for tagging changes.

bart's picture

I've made a couple of minor changes to the way art tagging works:

  • Spaces are now treated as breaks between tags, so you no longer have to separate them with commas.
  • Hashtags (#) are filtered out automatically.

At the moment, there are in excess of 11,500 different tags.  Of these, just about half are used exactly once, and about 800 aren't being used at all (most likely because they were attached to spammy art).  To really fix OGA's tagging system, we need to do a lot more:

  • Find all of the misspelled tags and replace them with their correct spellings (this could be at least partially automated, although the replacements would have to be reviewed before they are applied).
  • Set up term hierarchies, for example:  tree -> maple, so that if you search for "tree", you'll get things tagged "maple".  This is actually an interesting example, because even though all trees are plants and it would be intuitive to put "plant" one level above that in the hierarchy, when most people search for the word "plant", they probably mean to exclude trees.  In other words, the hierarchies need to be built with what people want to search for in mind, as opposed to technical correctness.
  • Merge certain tags that mean exactly the same thing (such as "16x16", and "16 x 16").
  • Create synonym links between tags for cases where the meaning is slightly different, and then weight the synonyms lower than the literal term in the search results.  For instance, if a user searches for "tree", it would make sense to give them results tagged "trees", and vice-versa.
  • Create a user tagging system that would allow users to tag submissions, and then upvote or downvote tags that other users have added.
  • Create an admin interface so that new tags can be curated as they are added.  Particularly if users are allowed to tag art that's not created by them, we'll have to keep an eye out for misuse of the tagging system (tags that comment about the quality of the art rather than the content).  Since we already have a user reputation system in the form of points, perhaps we could award user tagging access to people who have reached a certain point theshold.
  • Create a better tagging widget with type ahead find that shows tag hierarchies and related items, which will encourage people to assign their art to existing tags. 
  • Build a "popular tags" view.

I don't want to consider any of these items to be official TODOs just yet.  I'd be interested in hearing comments and suggestions from the community about how to proceed on this.

Please post your thoughts here.