We have tons of synonyms, now what?

Question

I just deployed some improvements to the tag synonym page

1. You now see a count on the target if it happens to be lower than the source (very bad)
2. You now see a count on the source if it is larger than one
3. Lots of ways to sort it
4. Mods can quickly blow away bad synonyms from that UI.

Now that we are clearly highlighting the issues how do we clean up the system and merge tags?

1. Should we post new retag requests for the tags that need to be merged?
2. Should mods go ahead and merge vetted synonyms (like iphone-sdk -> iphone) - keep in mind we have no undo
3. Should we automate part of this process? If so how?

Any other ideas? How do we cope with tag merges after synonyms are created?

Answer 1

Are there any plans for instructing people to create a tag synonym if they try to create a new tag that's not already a synonym?

Right now, I could create new tags as I see fit if they aren't a synonym in the system. It would be nice perhaps to see a message that suggests I look into creating a synonym or use the parent tag instead.

Answer 2

Right now there are 6 pages of 100 tags. That is quite a few to look through all at once.

It would be nice to see a filter or search for this page. You could filter by #of retags, last retag date, cases where the source has a higher count than the target.

Or, you could let people search by tag names.

Answer 3

After a synonym suggestion gets the required votes to become an actual synonym, there should be an automatic mass re-tagging of all sources to the target.

That being said, I think 3 votes to 'vette' a tag synonym is too few.

Answer 4

Adding a synonym should retag existing questions. Otherwise, what's the point?

Edit: In light of the recent bad synonyms being put into the system, I think auto-retagging might be a bad idea. Perhaps there could be a moderator-only system where if a synonym proposal meets quorum, the proposal gets flagged and a mod could allow it (the system would then auto-retag) or deny it (and dump the proposal) with one button click.

To counter the problem of low-traffic tags, I think the thresholds could be based on the number of questions in the (target tag + source tag) set for the past month or so. Something like the tiered thresholds for the vote-to-delete system, but with a cap, because the question counts usually only increase over time.

How about a minimum of 3 votes plus 1 extra vote for each ~2,000 questions (per month) with a maximum of 10 votes.

I think this strategy should be in addition to the only-users-with-upvotes-in-the-target-tag-can-propose-synonyms idea.

---

This gets the question count for the C# tag for June 2010:

SELECT COUNT(*)
    FROM Posts
    INNER JOIN PostTags ON PostTags.PostId = Posts.Id AND PostTags.TagId = 1
    WHERE
        Posts.PostTypeId = 1 AND
        CreationDate >= 'Jun 1, 2010 12:00' AND
        CreationDate < 'July 1, 2010 12:00'

This returned 6470, or ~6 votes to create a synonym.