Google Blogoscoped


Forty Faces Back-end  (View post)

Utills [PersonRank 10]

Sunday, November 20, 2005
16 years ago

How about some sort of picture cloud (like a tag cloud)...which would maybe try to highlight frequent posters.

You can give scores to different authors by links to site, pagerank, average comments per post...etc...and then multiply by how frequent/fresh the post is and order the pictures into a cloud.

You'd need to keep it good enough so that the good blogs dont deny smaller posters from getting noticed but at the same time it could mean that people who actually are interested in the bigger posters can see straight away the fresh posts by a popular blogger.

You could perhaps also fade the pictures as time passes between posts...perhaps on client side (if this is possible in javascript) that the picture cloud stays useful.

On a related you poll the RSS even in the night? or do you only poll during the day?

Justin Pfister [PersonRank 10]

16 years ago #

Nice Job and well documented. I like the tag cloud Idea. Yahoo's API for Content Analysis might be fun to use. But the Term Extraction service is limited to 5,000 queries per IP address per day. Maybe you could calculate each bloggers most clicked on article and then do Term Extraction on it. That would allow you to support 5,000 authors. Here's a link :

alek [PersonRank 10]

16 years ago #

Interesting post – always nice to hear how the behind-the-scenes work. On your "go from 1/2 hour to 4 hours" polling stuff, you might consider an exponential backup – i.e. 0.5 hours, 1 hour, 2 hours, 4, 8, 16, 32, etc up to something like a week (or maybe even higher). This way, blogs that go dead (really) won't get pinged even every 4 hours ... but if they finally DO come back to life, you'll eventually pick up on 'em and reset your polling interval.

Cem Basman [PersonRank 1]

16 years ago #

Thank you for looking us under the hood. In the further refinements please stay clean in your concepts. That's the strenght of your solution.

Clean concept, clear and robust realization, nicely documented. I like the simple idea. It has a special quality.

Philipp Lenssen [PersonRank 10]

16 years ago #

Utills, your ideas are good and got me thinking. Checking the "authority" of the blogger might be reasonable, but I wouldn't want to be yet another site which increases the authority of A-list bloggers. Because Forty Faces is a little bit about discovering blogs you didn't know of. But even measuring click-throughs or any other kind of non-random "FacesRank" has the problem that people who just blogged who then look for their face might not find it. And that would feel a little "broken" or at least take away from the fun of the site for those bloggers who joined...

Alek, the exponential "penalties" were my first idea, actually. It does make the most sense when considering a server is down forever (actually, I'll probably just completely stop the feed when the happens). But why I didn't do this is because when I would be able to pick up a feed again, my estimate as to the new post's age is not very good – it could be 5 days old or so, and wouldn't fit on the "just blogged" idea. Of course, I could check the actual RSS time stamp, but I don't want to get into trusting those :)

Cem, absolutely + completely agreed. The whole idea is "visual, simplicity, human (faces)". There's no Ajax, no tag clouds, not even client-side JavaScript (if we disregard the Google Analytics stats counter). If I can't find a way to enhance the site with keeping that approach of simplicity, I might as well live with the fact that the 40 faces on front will not show everyone everytime...

Artem [PersonRank 4]

16 years ago #

How about using the blog popularity for ranking the faces?
AFAIK, currently whoever blogged last is shown upper.
What is you rank faces using formula similar to minutes_since_the_last_update + blog_popularity? Whoever gets more points, is shown first. this would give a chance for any blog to be shown up (if a person posted just before the update), but chances of the popular blogs would be higher.

By blog_popularity I mean NOT PR, but the popularity among the fortyfaces users. For example, it could be ratio of:
number_of_clicks_on_the_face / area_occupied_by_the_face.

Let me give an example to make it a bit more clear.
1. Imagine that someone accidentally clicked on my face and was really impressed by a chance to propose a name for a real Google Bowling Team (
2. Let's imagine :) that the post was so successful, that a lot of people discussed it and many are looking for my face on the
3. Let's then assume that I don't blog for a long time and much later publish the googler's opinion on Amazon Prime algorithms (
4. Everybody will click this my face (because everybody was waiting for it)
5. I get a real big popularity rate, because even though my face appeared just once per month, everybody clicked it

I hope, I managed explaining the idea :)

Of course tracking clicks means, that you'd have to use some kind of redirects instead of direct links

Kyle [PersonRank 0]

16 years ago #

Or you could simply use a javascript onclick event for the link that increments the counter. But that'd probably require the use of ajax

John K [PersonRank 2]

16 years ago #

Maybe if Google does buy Riya, you can hook into that, and not worry about the table that matches the name to the face :)

Of course, Google doesn't usually hand out APIs...

Forum home


Blog  |  Forum     more >> Archive | Feed | Google's blogs | About


This site unofficially covers Google™ and more with some rights reserved. Join our forum!