11

Possible Duplicate:
Migrate shortlist based on tags?

I am pretty active on the R tag on SO. It's a language designed for statistical programming, so we wind up with a lot of questions that are off-topic for the R tag but on-topic for CrossValidated.

Here's the rub. When I vote to close a question as off-topic, the system suggests:

suggested migrations

Of those, only meta and TeX are likely to see (or have seen in the past) substantial numbers of migrations. And CV is by far a more common migration than either and is nowhere on the list.

Would it be possible to have the suggestions be data-driven on a per-tag basis? E.g. it suggests based on where migrations from a tag in the past have been sent to.

  • The list of migration targets has recently been updated, according to frequency of (successful) migrations. I doubt CV is really more common than either of these. – Daniel Fischer Jul 2 '12 at 21:21
  • 3
    I like the idea of making that context sensitive though. – Flexo Jul 2 '12 at 21:25
  • Who defines the context though? It means maintaining lists of sites that might be pertinent to different tags (I assume tags would be easiest). – ben is uǝq backwards Jul 2 '12 at 21:27
  • 1
    You could pretty easily build a simple predictive model that guesses the most likely migration targets for a given set of tags. Use whatever method you want, from simple regression to fancier stuff. – joran Jul 2 '12 at 21:29
  • 1
    The only problem I have with this is most questions asked on the wrong site (I'd say more than 90%) are not good enough to be migration-worthy. Migration is the exception, not the rule. – user102937 Jul 2 '12 at 21:43
  • @RobertHarvey It makes suggestions currently; they're just not as good as they could be. And the data input into the predictive model would be based on successful migrations in the past (e.g. your data is high quality). – Ari B. Friedman Jul 2 '12 at 21:50
  • What I mean is, a predictive model like this could encourage migrations of questions that shouldn't be migrated at all. The predictive model is based on tags, not question quality. – user102937 Jul 2 '12 at 21:51
  • @RobertHarvey I'm confused. The model I speak of would only control which sites appear on that list, not the actual migrations themselves. – joran Jul 2 '12 at 21:57
  • @joran: Yeah, I know. The reason Programmers.SE was removed from the list (even though it was one of the top 5) was because it was getting too many crap migrations. – user102937 Jul 2 '12 at 21:58
  • This question takes a different strategy which tries to drive people to ask the question in a different place to begin with – Some Helpful Commenter Jul 2 '12 at 22:04
  • 1
    @RobertHarvey Perhaps your concern could be addressed one way or the other with some data looking at whether there's been a significant increase in the number of rejected or low quality migrations to tex since the change? – joran Jul 2 '12 at 22:10
  • @SomeHelpfulCommenter Good find. I guess two things make this distinct: 1) The predictive model rather than just a hard-coded list as that question implies, and 2) being a statistics-driven person, I'd say the whole list should vary :-). Still, probably not enough difference to make this not a duplicate. – Ari B. Friedman Jul 2 '12 at 22:12
  • What about a user specific migration list? Is it possible to see which stackexchange sites a user frequents and perhaps even their activity? Most active sites could come up in the list for migration. – Roman Luštrik Jul 3 '12 at 8:54

Browse other questions tagged .