June 02, 2008
People Augmented Search Engines
In Search 4.0: Putting Humans Back In Search , Danny Sullivan comments "I think there is some potential to tapping into a social network and applying it to search. However, I still remains uncertain how that will unfold. It especially remains uncertain that this is somehow the secret sauce for anyone to jump past the current state of search."
I've been thinking a bit about that lately, as another piece in a jigsaw I don't quite understand (although it does seem to include elements of serendipitwitterous and the social life of a URL (of which, more in a later post)).
I've also been thinking about things like this:
Firstly - tweets appearing asking friends/followers a question; secondly, the appearance of (advertised) services that are 'human-question answering service' like...
So here's what I'm thinking... if the search engine knows who your friends are, then when you post a query, it sends the query to them, and they may or may not answer back. It's unlikely that they'll respond quickly enough to make it into a favoured 'Your social network suggests' results slot at the top of the results listing, but friends' results could appear in a sidebar, in effect providing a lazy, slow channel results listing, but hopefully one that is very high quality, and/or maybe usefully related to your recent queries?
This model in turn becomes a driver for natural language search (and the construction of longer, more specific search queries) - because you're actually asking people wo help, as well as the search engine.
Alternatively, if the search engine knows what sort of people are the sorts of people who might know the answer to your question (maybe because the way they have been 'indexed' serves them up as a highly ranked 'result' when you make a query), then it will add the query to their stream, and if they answer, the result comes back to you...
(Obviously you'd want to add some sort of load balancing to make sure the same people didn't get bombarded with search requests).
In the event of a likely "person who might know" not being online, then potentially a search could be made over their social bookmarks, on the grounds that these are already 'implicitly recommended' by virtue of having been valued enough for them to be bookmarked?
So the protocol is:
1) I make a natural language search query;
2) the search engine does its normal thing and brings back its normal results;
3) the query is streamed to my friends;
4) it may also be streamed to people who are ranked highly when this search is run against their indexed profile ('people who might know');
5) if a friend replies, it appears in your lazy results sidebar;
6) if you friends/'people who might know' have a social bookmark collection, the query is run against those results, and if any relevant ones turn up then they may be offered as 'x might suggest' highlighted results at the top of the results listing.
7) if the user has personal web history/search history tracking enabled, and any of these links turn up as highly favoured, they could be returned as "you may be trying to remember" result listings?
Hmmm...
PS this just in: @answerme Answers Your Questions On Twitter
Tags: twitter, search engine, social search, human powered search
Posted by ajh59 at June 2, 2008 02:27 PM