You are viewing a single comment's thread from:

RE: Programming Diary #19: Follower network strength and preparing for Open Source publication

in Steem Dev2 months ago

Interesting. That even predates me by a a couple of months. I have no recollection of ever seeing that feature in operation, so maybe they removed it quickly. If I remember, I'll ask @cmp2020 if he remembers it when he gets home from college. I didn't find any other references to it in github or google after a couple quick searches.

I agree that it was likely removed due to performance considerations. That definitely strikes me like it would have trouble scaling. Also, when the automated voting came along, I'm not sure how accurate it would be. I think useful recommendations would probably require something more sophisticated (i.e. language processing & machine learning). There's a whole field of AI research that's dedicated to this purpose.

If I scroll to the bottom of steemitblog, I get there. If I scroll to the bottom of steemitblog/posts, it ends prematurely. Both go all the way to the bottom from steemitdev. Maybe another bug that you fixed with your work on condenser. 😉

Sort:  
 2 months ago 

I think it was the first post that came from the steemitblog account. There was a latter post about Steemit going open source so my suspicion is that the code didn't even hit GitHub. I haven't really looked though.

I think useful recommendations would probably require something more sophisticated (i.e. language processing & machine learning). There's a whole field of AI research that's dedicated to this purpose.

Good point. I suspect that it won't be long before an AI model can analyze a user's posts, review who they vote and comment on and subsequently find authors with a similar profile. I would say that using tags (for example) but few people use tags correctly now. Ignoring that downside, it actually feels like a fairly straightforward algorithm to do.

  1. Review my comments and posts.
  2. Review the people I've commented on (tags used, reputation, date joined, etc.) with pre-determined weightings.
  3. Find people who use those tags, have that reputation, etc. with a narrow to broadening search. It could target 3 months data so as not to be overwhelming.

Maybe another bug that you fixed with your work on condenser. 😉

My guess - it's hard coded on the steemit.com domain 😉