Thanks for the shout-out on my Re-Thinking Curation post. I feel that there's a ton of articles that are being overlooked because the users simply don't have enough followers / reputation.
There's also a massive diversity problem. Most of SteemIt loves crypto and talk about the SteemIt platform. These categories are likely to be disproportionately rewarded because they have larger audiences and more whales.
Conversely, a category like pregnancy would likely attract very few high SP users and a relatively small audience.
I'd be curious to see some statistics on the most commonly used tags among your "underrated posts" - I wouldn't be surprised to see crypto-related content filling much of the top end of your graph.
Yes, diversity will be a problem, maybe I could look into the truffles for different tags. I have to see whether it even make sense to train different models on different tags (yet in this case the data might become quite sparse, especially for rare tags like
pregnancy).You're right that this isn't very useful for obscure categories with very few posts (#belegarth?) but it might prove effective for evaluating relative value among popular tags relative to other posts in the same tag. There's a list of popular tags here which might be a good starting point.
It would be a really nice metric to be able to say something like:
If you haven't already, I'd recommend checking out @thing-2 and @gentlebot - They're both comment evaluator bots that try to do something very similar to what you're doing with this bot.
Thanks for the tip, I'll look at them. And I think about post statistics for different tags.