Posts Tagged ‘news’

Personalized search from Microsoft

May 5, 2010

Lili Cheng of Microsoft FUSE labs has been sharing with the web2.0 world their newest service, SPINDEX.  The goal for the services?  To “.  .help you get the most out of your social activity by exposing the right information, at the right time, in a way that’s meaningful.”

WOW, they don’t do small visions at Microsoft.  I think when you combine the right information and at the right time you are framing context.   There is presentation context and authoring context, really two sides of the same coin.  Its search that brings (Bings) the two contexts together.  Not being in SF this week, I did not get an access token to their site but my expectations are for an activity stream UI with some ‘filtering’ UI tools to make it different from e.g. a twitter or FB news feeds.  Lili concludes her blog post by saying, “There’s still much territory to be surveyed. . ” in these early days on the web.  I agree, this context awareness activity stream web will change the face of search.

Advertisements

filtering social news

April 23, 2010

What are the magic ingredients that brings the best information to us in a social context?  A couple of months ago I presented the whole story on the logic going on behind mepath.com.  This week the development team over at facebook partially presented their model.  The techcrunch blog has a great summary.

EdgeRank is the core metric formed and a news feed is a series of Objects prioritied by edgescore.  This can be summarised in the following equation:

The Sum of all Edges = Ue * We * De

where U is the affinity score between the user and Edge creator, W weight for each edge type, D time discount.

U – is real interesting as it is trying to decide how close you, the reader of the news stream is to author of the object.  Now, FB has a great starting point, they know there is a social connection between you and the author as only friend connections are brought back into a new feeds (OK, slightly more complicated as fan pages etc or app. data may feedback object not authored by a friends ie. like a RT of a non follower on Twitter).   By contrast mepath is given no starting relationship connection, it has to decide that itself and does so based on the lifestyle context of an individual over time.

W – applies a weight based on the type of edge e.g. comment, video post etc.  In effect saying the type of media authored can be biased up or down.  That is real interesting but how these weighting are decided is of even more interest?

D – time discounting.  Real time to later time.  Everyone one wants the latest but just in case we are not logged in 24/7 then we can review top news that ‘holds’ on to some friends post for longer, minimizing the chance you missed important news.  From my thinking on time filtering data, it is a deep and complex science all on its own.  The most important information an individual wants may well have happened in the past and well into the past.

While we can segment and split all we like, the simple objective remains, bring back the best information that is possible to have for the individual.  And increasing the will found and be achieved by understanding both the context it is initiated and understanding the context in which is was authored.