Max's Output

Archive for November 2009

Extending The Twitter with Source-Oriented Algorithms

There are several ideas on what next significant features can be added to The Twitter In this post I summarize a number of my favorite ones that are derivatives from the following general idea: using Twitter as a voting system with respect to a source (e.g. a blog feed or a Twitter user timeline) or a bundle of sources (e.g. Google Reader bundle or a Twitter List).

How it works

It works as follows. Links published by a source are ranked or filtered with respect to how many times they are posted on Twitter. For example, if the source is your favorite blog, which can be too fruitful to read all posts though, such a system allows you to identify interesting posts that you should not miss in the blog.
In case of ranking, it works like The Twitter Times: links are just ordered according to their popularity and recency on Twitter. There can be two reasonable options when it computes popularity. One is to consider only tweets from your friends and friends-of-friends (fofs) as The Twitter Times does. Another is to consider all Twitter users (still with friends counted with a higher weight) because for some sources there might be only a few or no tweets from your friends.
In case of filtering, the system identifies a subset of the source’s links that should interesting for you. It can be implemented by selecting those links which popularity is above the average for the source. As in case of ranking, it is very likely that we will have to consider all Twitter users (not only friends and fofs) to implement the filtering.
To conclude I summarize the difference between friends-oriented ranking (currently supported in The Twitter Times) and source-oriented ranking/filtering described above:
1) The friends-oriented algorithm uses Twitter as a voting system for all the links posted by user’s friends and fofs. The source-oriented algorithms consider only links coming from a single source.
2) The friends-oriented algorithm counts votes only from user’s friends and fofs. The source-based algorithms have an option to count votes from all Twitter users.


There are at least two interesting applications of the source-based ranking/filtering algorithms.
First, it can be used to identify interesting posts from your favorite blog, bundle of blogs, or newspaper (for example, The New York Times or The Los Angeles Times). Each day (or hour) you can read a ranking of posts or a list of selected posts for your favorite source which are hot on Twitter.

Second, it can be used to build thematic newspapers described by Maria Grineva in her post. You should just create a bundle of blogs (using for example Google Reader bundles) or a Twitter list which includes sources covering a common theme.

In both cases it can be embedded in The Twitter Times interface as a side bar or a tab for each source (bundle, Twitter list).

Written by maxgrinev

November 22, 2009 at 5:45 pm