Social Graph Search Engines, Part 3
Posted by Bob Warfield on August 27, 2007
I’ve inadvertantly stumbled on a topic that’s getting tremendous attention, so I wanted to sum up my thoughts here before moving on.
Let’s start with Robert Scoble’s post, Why Mahalo, TechMeme, and Facebook are going to kick Google’s butt in four years. Scoble’s proposition, which he communicates well in three videos, is that by using the Social Graph, we can dramatically reduce SEO Spam in search results. The source of the Spam is that Google’s current PageRank algorithm is just too easy to game. You simply need the right keywords together with as many sites linking to the page with the keywords as possible. That combination does not guarantee a human will see anything of value on the resulting page. With a Social Graph Search, you will weigh in human judgement about the quality of the page. You do this by considering factors such as:
- Did someone from your Trust Fabric (i.e. your social network) write the page?
- Was it written or linked to by people who you know share similar interests because of the social graph?
And so on. It’s a good idea, and I had already modified my own search habits to take advantage of a less grandiose form of this by choosing to search blogs before doing a general Internet-wide Google.
If you read through the comments on Scoble’s post, there seem to be two big issues with this proposition. First, its simply very hard to get over the idea that Google is huge, omnipotent, and can simply tweak their algorithms to adopt a Social Graph Search approach. Second, there is a lot of doubt about the feasibility of human augmented search.
Let’s start with the feasibility of human augmented search. I think people are looking at it too literally. Perhaps companies like Mahalo are also in the camp of being too literal. If we have to wait for humans to laboriously research each topic and select the best links in some systematic way, I agree that sounds painful. Not only that, but it seems like its been done. Isn’t that what About.com did? I’ve landed on their site only rarely and never found it to be helpful. But look at it another way. I do a lot of research on the web, and there’s nothing better than hitting pay dirt by finding someone list of bookmarks are a link to the seminal article that started some new idea. Today, I find these things by tracking through blogs, and its a lot faster than tracking through Google. Couldn’t that be automated?
So in a nutshell, my argument about feasibility is that there are already structures in the Web that are mostly Spam-free that can be tapped into to deliver improved search. The blogosphere is one such. I can already see an objection: won’t the SEO’s just shift their attention to polluting these new information watering holes with their Spam? Perhaps, but many of them are more amenable to Spam defenses because they’re not just randomly selected web pages. Let me give you an example. Let’s say we’re going to harness bookmarks in the blogosphere, for example. Let’s have the human editorial staff simply identify a core nucleus of trusted blogs by reading them. We’ll take blogs like Scobleizer, TechCrunch, and Mashable. Let’s tag pages on those sites with a “Bob” Rank (because hey, the Page Rank is named after Larry Page!). The BobRank for pages our editorial staff have validated is infinite (in Computer Science, that means its a really big number, like 1 billion). Every link on a 1 billion BobRank page gets a rank of 1 billion minus 1. Links on those pages are 1 billion minus 1 minus 1. And so on. The BobRank is telling us how far away from objectively validated acceptibility a page is.
We can envision many embellishments. For example, we can augment the BobRank with a FriendRank. Have my friends in my Social Graph written or bookmarked the page? Well then up the BobRank by 1 million. Have their friends? Up it by 1 million minus 1. And once again, we have a super Page Rank-style algorithm where all the people in the Social Graph are helping out.
How do we efficiently gather and process all this information? That brings me to the first objection, that Google is so big it can just do this itself. That’s very true, BTW. But it doesn’t mean everyone else has to lose out. The rest of the AltSearch community simply need a Queen from which to launch Swarm Competition. I’ve already mentioned how this could be done around Yahoo. Let’s revisit in light of the Social Graph-based strategy how Yahoo can jump about 2 generations ahead.
First, Yahoo owns enough properties (such as del.icio.us) that they have Social Graphs out the wazoo to analyze and create a “Jerry” Rank (they certainly won’t call it the “Bob” Rank, doh!). But secondly, they can open up their world and leverage the AltSearch community of something like 1,000 different search startups. My proposal was that they turn search itself into a Social Network. They can do that by leveraging all these social graphs, by releasing their own open Identity and Trust Fabric (e.g. Social Graph) apis that their partners must share to participate, and they can make available their web crawling data and massively parallel horsepower via Hadoop.
Okay, we see how Yahoo benefits. Big hype value. Big news value. Promulgating a key open standard for Identity and Trust Fabric. Leveraging everyone else’s social graphs to improve their search and generally creating network effects that haven’t been seen since eBay rolled up the auction world. How do the partners benefit?
To continue the idea of building a Search Engine around a Social Network, what else do social networks have? Widgeits! Look how excited everyone got about Facebook Widgets! Yahoo can bestow on their partners the ability to create search widgets. People in the Search Network get to choose which widgets they like. From a UI standpoint, it might not be too different from how Microsoft shows different search views in Tafiti. Remember the little ring with the icons? Or it could be a cleaner textier Google, sorry, definitely it would be a Yahoo-style list. What’s on the list is determined by what widgets you signed up for, but Yahoo will also give you a basic set for general purposes.
To this we also add feeds, photos, and all the rest of the stuff that makes up a Social Network.
Still think someone can’t get there with a decent start? This vision is my bet for what search engines look like in a couple of years, as well as an example of something I’ll talk more about:
Any web offering can be turned into a Social Network, and there are huge Competitive Advantages to doing so.