I’m thinking about possibility to make more sophisticated SE data dumps analysis in order to select most interesting questions. Maybe not a problem for SO because huge number of questions have 3-digits votes counts. So big numbers work very well here and for every adequate tag a novice can find a lot of questions worth reading. But maybe for smaller sites/less popular tags/… the existing very simple voting system could be boosted?
Ideas I have in mind are
- using of reputation of people who participated in the discussion,
- using of average votes for questions/answers posted by each participant,
- using of votes to views ratio (or additions to favorites to views ratio),
- to collect as much characteristics as I can invent and to feed them to a data mining program. In order to find correlation of some combination of them with vote/favorite count,
- some iterative algorithm, e.g. user with high rating give some points to questions they answered, questions give some points to all who participated in them, and again from the beginning.
What do you think, is it worth trying? Which other algorithms could you propose?
"reputation of people"
- Probably not, I've seen plenty of 100k+ users comment on / answer terrible questions. Participation doesn't mean you think it's a great question.