Batch retagging?

Question

What is the recommended way of dealing with retagging many questions?

For example, if there are only 6 questions that use the smething when they should be using something, it's easy enough to manually retag them, although you get the captcha every two edits, which it rather tedious.

This doesn't scale very well though - for example, something that bothered me is there are about 300 questions tagged as subversion and about 1,500 tagged as svn.

Obviously mass-retagging could really break things, so it shouldn't be a one-click oops-all-.net-questions-are-now-tagged-subjective.

Perhaps it could be based on voting (in the Tools menu), with a moderators final approval?

Answer 1

Good news

This is now available as a moderator-only function, thanks (as always) to Jeff and the team.

So; if you have a need to merge two (or more) tags, please propose it here on "meta" (so it can be discussed etc), stating:

• which site
• which master tag (the one to keep)
• which tag or tags to migrate and delete

and if appropriate it will enacted by a moderator

To reduce the scope for error, I would anticipate:

• a grace period (to allow discussion/disagreement along the "no they're very different" lines)
• a question back on the affected site (tagged with all the tags involved) along the lines of "tag-merge proposed [link]: are these the same thing?", so that the people who know those areas have visibility (the moderators can't necessarily be expected to know the area(s) well enough to directly approve/veto the merge)

Answer 2

I'd like to see a way that you could nominate a tag for elimination or merging into another tag. Then it would show up in the 10K tools section where those users could vote for or against the nomination, and hopefully also post comments or concerns. Enough votes and it all happens at once over night.

There are some issues with this, though:

• It's tough to automate tag elimination, because sometimes the tag you want to eliminate is the only tag on a question. We need to know what to do in that scenario.
• Do we want to restrict the future use of eliminated/merged tags?
• When merging tags, we need to add the merged tag to the job that automatically updates tag synonyms.
• When merging tags, how do you handle multiple distinct synonyms. For example, correcting 'access' to become 'ms-access', what about questions pertaining to "disk access"? Obviously those latter questions are tagged poorly as well, but that doesn't mean they won't exist. Perhaps that becomes part of the discussion, where the "disk access" questions are identified and corrected manually first.

Answer 3

Another issue with the 1-by-1 approach (if you take the time) is that you swamp the front page... suggest that batch retags shouldn't set the "last updated" (or whatever) that triggers the front page.

Answer 4

I set the # of questions per page to 15, open the first page worth of questions in browser tabs, retag each one (closing the tabs as i go), refresh the tag query page, and repeat.

I don't do this very often. It's a huge PitA. And as Marc notes, anyone doing it this way pretty much owns the front page 'till they're done.

A system where you could propose a tag merger, comment on it and vote for or against it... would be great. Heck, we could actually use this Meta-SO site (via questions with a retag-proposal tag) for that purpose, if the admins were willing and able to process the results...

Answer 5

The problem with this is it won't always work, I for example moved from rosetta to rosetta-stone, because there already exists something with that name, that has nothing to do with a question that asks for examples in other languages. Which means the tag should still exist, but many thing that were tagged rosetta had nothing to do with that technology/module whatever.

Answer 6

I think that rather than attempting tag mergers there could maybe be a way of creating tag synonyms.

For example 't-sql' and 'transact-sql' are made to be synonyms for 'tsql'.

Otherwise you'll have about as much luck as King Canute...

Answer 7

Automatic changing of tags can be very problematic. I think that re-tagging really needs to be a manual effort because you have to read the post to figure out what the tags should be.

However, the re-tagging process needs some serious streamlining. I occassionally try to do some re-tagging because I have slight OCD when it comes to organization of things, but it becomes incredibly tedious in a short amount of time.

What we need is system that will support re-tagging in batches:

Step 1: Select Source tag Step 2: Select Target tag Step 3: System shows you posts with the Source tag one at a time. For each post you say "Source", "Target", "Other", or "Remove" to choose whether to leave the Source tag alone, convert it to the Target tag, convert it to some other tag, or remove the tag entirely.

This would make it very easy to convert tags but allow manual confirmation that the tag should be converted. I don't think that it's possible to fully automate this process but the process described above would make things as easy as possible to correct.