Over the past day there has been significant discussion about what the future holds for real-time search, and no matter what conversation you look at, Facebook is always part of it. Just last month, Facebook started testing a new version of search and over the past few weeks, I’ve taken the time to play with the system and compare it to other “real-time search engines”, primarily Twitter search.

Twitter Has More Volume, More Noise

While Twitter appears to be leading the real-time search engine, real-time search is far from mature. We’ve only begun to define what real-time search means but the assumption is that real-time information can add relevance. Since you can tweet information quickly, we can get more relevant information quicker. For example I first learned about Michael Jackson’s death a couple weeks ago via Twitter and then read about the confirmations and also received confirmation from a close source to the pop mogul.

Twitter ended up being extremely relevant in that instance and there are many other instances in which this information can be relevant. Twitter isn’t the end all be all of real-time search though. While there are currently a number of filters via Twitter’s advanced search tool, there is still a large amount of excess noise. I don’t know most people on Twitter and reading through every tweet in a search query, sorted by date/time doesn’t make a lot of sense.

This unfiltered access to all tweets can be seen as an advantage or disadvantage of Twitter search depending on how you look at it.

Facebook Uses Structured Data To Improve Relevance

If you asked me how to improve relevance in the world of real-time search, I would argue that by using data about the person searching, you can dramatically improve relevance. Luckily for Facebook, they already have a number of variables, many of which I outlined in January, that can be used to increase relevance. Here are the ones I mentioned at the time:

• Comments on shared stories
• “Friend Rank” - A way of valuing the relevance of a link based on a user’s authority via the number of friends they have and the influence of each of those friends
• Number of times a story is shared
• Networks users are members of (good for targeted search)

There are a number of other variables that could be used but the point is that each variable can be used to increase “relevance” and ultimate that’s what the real-time search war is about. So far all that Facebook has provided for searched items is feed postings by your friends as well as all users that have selected the “everyone” privacy setting for posted items. Currently the volume of public items is well below Twitter but honestly, volume does not equal value.

Phase Two of “Real-Time” Search

We now have a the “real-time firehose” of data from Twitter, Facebook, and FriendFeed (as Danny Sullivan highlighted yesterday). Now we need to make that information more useful. Facebook can use their own custom “relevance algorithms” to determine which feed stories are most relevant. For those stories that have been shared privately, Facebook can simply find the one individual that shared the story publicly and use that in the results.

I’m now getting in to the technicalities of how it would work but the main point is that Facebook is at the beginning of their search evolution. The next phase involves relevance and Facebook is well positioned to provide that relevance thanks to all of the personally relevant information that they have about us. It’s time for Facebook to organize the information coming out of the real-time social firehose.