aluhrs13 / TheMiniIndex

Crowd-sourced library of 3d models (minis, terrain, scatter, etc.) for D&D, Pathfinder, and other tabletop games.
https://www.TheMiniIndex.com/
28 stars 5 forks source link

Add "Preapproved" State #222

Open aluhrs13 opened 3 years ago

aluhrs13 commented 3 years ago

Problem - I try to only approve a handful of Minis per day to avoid flooding the homepage with a single creator's content. This ends up with me forgetting or being too busy some days and the homepage stagnates, or I just get too tired scrolling through the one dude who tagged 40 of his own Minis to find someone else's.

Goal - I'll "Preapprove" these Minis, and write a script or Azure Function that will change N things from preapproved to approved every M hours.

anaximander23 commented 3 years ago

What criteria are used to select which minis to preapprove? I looked at some ideas for autoapproval; I think we may be able to recongise minis that are likely to be approved by looking at historical stats. For example, a new mini by a creator with many approved minis and few or no rejected minis is likely to be approved, while a mini by a creator with few or no previous minis, or a creator with a higher ratio of rejected minis, is more likely to need manual review.

aluhrs13 commented 3 years ago

The problem isn't really the Mini's approval, it's the tag's approval. Your approach is/would be good for unindexed -> pending, but pending -> approved is approving more the tags than the mini itself and that's the one with the most overhead right now.

Now that I think about it though, the same concept might apply but to the tagger instead of the creator. There's a few people who are reliably "good" taggers and a few people who put a dozen unrelated tags that I have to remove before approving. Giving someone a reputation based off of number of removed tags might work, maybe I can write a script to get data to prove that...

anaximander23 commented 3 years ago

That makes sense. That ties into a concept I think we've discussed before, too, where we could look at the rate at which certain tags are approved and use that to identify combinations that are statistically likely to be acceptable. That might be that certain tags are usually correct, or that certain users are reliable, or some combination of those. A similar thing might be useful for auto-tagging; set up a process that applies tags in a pending state, track how often those tags are then approved or rejected, and auto-tag rules that have a high approval rate can transition to automatic approval.