Thursday, 17 January 2013

Facebook Graph a Semantic Social Search Engine

Facebook graph search has landed and with it a worlds first, a semantic social search engine. The vast quantity of work we have all put into Facebook, all those likes and status updates, the affirming and rejecting of friends. The location updates, the invites and group events. They all add up to an immense pile of neatly structured data. More than that. Data that is inherently understandable by humans. Compare that to the mountains of data utility companies and banks hold on us.  This stuff is about us, what we do and like and who we do it with. We all ways knew it was a goldmine, but now we can see how Facebook intends to profit.

The data is already ripe for semantic search. We have fed it data in social formats which relate to our lives and the times of our lives. So providing a social search engine is a brilliant thing to do.

Initially I though the data would be fairly crude in terms of how to rank and apply relevance to results. Facebook appears to have a  a very narrow tagging system you can either like something, comment or share. But beyond that there is how you structure your feeds, who you elect to receive updates from. If you read those updates, if you follow links. Once you add these variables into the mix they have a the ability to prove more granular or accurate results. I can't tell if they are doing this yet but it would seem the obvious path to take.

The auto complete in the search box offers terms which fit with the content types and relationships you are searching for. So  "National parks my friends.." could complete to "National parks my friends like" or "...have been to". It knows that national parks are places you visit, it will also know that restaurants are places you eat, Sony makes things you buy, Justin Timberlake is someone who makes music and has concerts you can go to, and that you can like and comment and share all of these.

If they open up to other services such as retailers, location based services -  FourSquare, brand and product recommendation engines, then the value of the Facebook data is released every time a user searches. First pull in related results from the external sources. Then charge to allow other sites to pull in social graph results into their results. Of course this will bring privacy issues to the fore again. But int he past Facebook has managed to ride the waves of discontent and come out winning.

1 comment:

  1. Interesting - but are you not familiar with Edgerank?

    I suspect that the same algorithm which influences what you see on FB (based on the degree you've interacted with a person or page) will influence search results.

    I'd also say this is different from semantic search - which I understand to relate to the comprehension of search queries - its personalised search of relevant to each user.

    Effectively combine the two though and then you have the killer search product!