Hot on the chronolog - and how it works


When I first published my chronolog, a few people were making remarks about how it resembles FriendFeed, Twitter or Tumblr. I can't deny that. The influences of Web 2.0 are huge both on my personal and business life, so why should the chronolog be any different? It is a mashup of different web services and it displays information from different sources, so it's a kind of a Web 2.0 stream. But besides that, it's also my own personal playground for testing and developing high level services and functionalities, which will hopefully be cool and fun and make the chronolog interesting for all of us. Demonstration of concept and technology, if you like.

I already have a few of those smart features planned, and I can give you a little teaser already. I really look forward to developing the custom view of the chronolog, where advanced users will be able to do a bit of configuration. The prototype is already half developed, but sadly far from production. A different thing I'm working on is a complex set of statistics and analytics, which should give us deeper insight into the chronolog, it's data and our interactions with it. This one will probably go out next and it actually inspired the one already complete. From this day forward, the chronolog supports Hot on the chronolog, accessible from the views menu top right, which shows the most interesting posts in the desired time period.

A few Web 2.0 portals (specially those oriented in social news or social bookmarking) have recommendation engines, which give users access to information based on their interaction with the system. I would like to try that one out too once, but because I don't have registered users, the chronolog probably won't be the environment. What I can give you now, is the Hot view, which displays the most important posts based on the interaction (views, likes, comments) of all users of the chronolog. A global recommendation engine of some sort. I'm quite pleased with the algorithm I've developed, it looks like it's working, so you can give it a try.

Some of you will be interested in how it works. The core is a really super mega awesomely complex algorithm that gives ponders to different interactions in the selected time span. Well, it's not that complex from the mathematical point of view, but it still pretty smart. Combining these ponders and number of interactions, using a few square roots and logarithms, plus a small modifier for insert date (if two posts are tied, the older one appears "stronger"), it calculates which posts are more interesting and relevant and gives them a score accordingly. Simple as that. Besides, is also able to make that calculation for any time period. You can even hack it by changing the ?d=# in the url to any number of days you like.

When viewing longer periods (months, years), blog posts will probably take most of the top spots, because they are supported with social networking sites and have the most interactions. In the shorter periods (days, weeks), other types of posts will also take higher ranks. We will see if the algorithm works in the longer term too, when more users will be clicking around, but if needed, the calculation will be changed or modified. Oh, I almost forgot about the design touch I added - the importance of a post is portrayed using transparency, which looks quite cool and is a great exaple of using design for function.

The chronolog becomes smart. Hope you like it.

A few more things you might find interesting:

written 10.10.2009 20:49 CET on chronolog
1694 views   •   3 likes   •   2 comments  •   Like   •