One of the biggest flaws of Google's algorithm is suggesting only known entities: you'll find already popular people and pages there. But some of the real gems in the G+ community are the folk who you've never heard about, because they don't have the same large following. +CircleCount 's new "explore" feature does that: It highlights interesting content, regardless of whether it comes from someone with 78 followers or someone with 100,000 followers. What a great way to find new people, and increase your general knowledge.
Hats off to +Daniel Sandstein and the rest of the amazing +CircleCount crew, for yet again filling in a big hole in what the G+ community needs!
Reshared post from +Jaana Nyström
What’s Hot in your country?
Starting today you can find some very hot google+ posts on +CircleCount/
► http://www.circlecount.com/explore/The main target of CircleCount is to let you find interesting people and posts which are not on the #SUL (=suggested user list; https://plus.google.com/getstarted/follow) or on the “What’s hot” section of Google+ (https://plus.google.com/explore).
That’s why the rules to get a post into this list is addicted to the number of followers. You can get even a post with only 16 reshares into this list.Another target of CircleCount is transparency.
That’s why you will find a clear description on this page, how you can get a post there.You can also filter these posts by the country or the continent where the author of the post is located.
Here are some examples:
► Finland
http://www.circlecount.com/fi/explore/► Germany
http://www.circlecount.com/de/explore/► United Kingdom
http://www.circlecount.com/gb/explore/► India
http://www.circlecount.com/in/explore/► United States
http://www.circlecount.com/us/explore/► Europe
http://www.circlecount.com/cont-eu/explore/Image credit: +Russell Deasley
This post has been reshared 1 times on Google+
View this post on Google+
+Abe Pectol thanks!
The post collection is running in the background and when a profile page is opened.
We'll check what we can in the future regarding language recognition (the easiest would be an attribute in the Google+ API with the language of a post 😉 ).
Consider your comments here as plusoned, since I'm not allowed to plusone it through the page 😉
+CircleCount you'll probably need it in some simplified and/or C-optimized form (it's easiest to do with Cython, but alas, you're using PHP 🙂 ). There seem to be at least something wrt C extensions for PHP… but not much (and not in most setups). And anyway, thinking a bit, it would probably be best to run it as a background cronjob (though, aren't you running the post collection process that way already?) that annotates post entries in the database with their language.
(sorry for a bit messy brain-dump 🙂 )
Also, a general note, from ML (Machine Learning, that is) perspective: what you do could be optimized (i.e. “made more precise & useful”) in general with some probability theory and as much computational resources as you have.
+Abe Pectol CircleCount is based on php/mysql. So the programming language would be php and if I find a (good) library there shouldn't be a problem to implement it. But we would have to check how the performance will be if "calculate" the language for each post.
+CircleCount there are already various implementations (and, maybe, libraries) for language detection/identification, so the point is to find what is possible (or “worth it”) at all given your resources (and programming language, and so on). I don't know either of those, of course 🙂
I would say both (dont know how and after thinking about it probably to complex?)
plus: too hardware-intensive (since each post would have to get analyzed)
But we are open for any idea/help! 😉
+CircleCount too computationally complex? Or just don't know how?
(even an approximation by used characters could be quite useful)
Unfortunately we can't use the language +Abe Pectol since we don't know the language. And identifying the language for each post is something we can't do.
Also, +CircleCount could use “[post] language” instead of “country”, I suspect.
M-mm, sounds like implementation (or re-invention) of “bayesian score” (… or however it's conventionally called).
Also, [bad] way of SaaS: most useful things has to be implemented as an external service.
Thanks for the flowers!
since you speak german, you know what I mean 😉