Aardvark began opening its social search engine to more users earlier this month, and now the ambitious company is starting to reveal more of its future plans. It’s bulking up its Google-heavy technical team with Sameer Paranjpye, the Director of Yahoo’s grid computing group and a founding member of the Hadoop project. It’s also taking aim at the mobile market, hiring on Ben Keighran, the founder of mobile messaging service BluePulse. I caught up with Keighran earlier today and he sketched out what sounds like a very useful iPhone application — it will let you get answers from friends wherever you are.
But hold on, what is Aardvark?
Put simply, it’s a new way to get answers from your friends. You email or instant-message your question to Aardvark, it figures out around half a dozen people you know who might have a good answer, then emails or IMs them for a response and sends what they say back to you. “Aardvark is supposed to work like a contact,” co-founder Max Ventilla tells me. “It should know you, understand what you want and make one-off introductions to the right person who can answer your question.”
What problem is this solving? Search engines are great for providing links to information that is online but sometimes a real conversation with a friend (or friend-of-friend) can provide much better information than a web page. After all, there’s far more knowledge and experience in people’s heads than there is written on web pages. Aardvark isn’t going to approach the utility of Google for information about things like “San Jose weather” or “history of computers” but it can connect you to the right person in your network if you’re looking for great live rock music or a tasteful restaurant for a date. In contrast to web search engines, Aardvark is trying to answer such subjective answers subjectively — a sort of real-time, personalized Yahoo Answers, and loosely similar to human-powered search engines like ChaCha.
How does Aardvark work? It analyzes social data about you and your friends, pulling in information like your email and chat contacts, and who your friends are on Facebook. If you give it access to Facebook — using the social network’s Connect service for exchanging social data with third-party sites — it will gather information from your and your friends’ profiles, like what geographic or work networks people are in and what their interests are. So, for example, if one of your friends on Facebook includes “action movies” in the interests section of their profile, they might be more likely to receive a question you send out about which Arnold Schwarzenegger hit you should rent tonight. Ventilla says about 40 percent of Facebook profiles contain enough information to be highly useful for Aardvark.
But Aardvark is trying to gather as much social data as possible, from everywhere. You can also directly provide Aardvark with keywords about your expertise, and when you invite friends you can tell Aardvark what you consider them to be experts in. And look for it to start tapping into other social networks that provide user data to third-party developers — the opening up of Facebook and other social networks over the last couple of years was one of the inspirations for starting the company, Ventilla tells me.
To be clear, though, social data is just one piece of what Aardvark does. The company, which is not shy about its engineering talent, uses machine learning software to figure out what people are intending to ask about when they ask a question — then it pairs that question with your friends. For a deeper look at how everything works, take a look at our write-up last fall.
Where Aardvark is going
In testing so far, around 70 percent of active Aardvark users respond to a targeted question, more than 90 percent of questions get answered — and over half of all questions get answered in under five minutes. But the results aren’t always perfect. The company is bringing in Paranjpye to help the service improve and scale. He’s a former search engineer at Yahoo — and before that he was a search engineer at Yahoo-acquired seminal search company Inktomi. In his most recent Yahoo position, he built up the 40 person Yahoo team working on grid computing initiatives and was one of the main architects of Hadoop: a software framework to build applications for processing thousands of petabytes of data. At Aardvark, he’ll be doing related work, helping to improve the speed and accuracy with which questions get paired with people.
Refining the core product is necessary for the company to start making money. For example, it’s already sending affiliate traffic to Amazon. The more successfully that questions can be paired with people to answer them, the more opportunities for this sort of revenue. Longer term, the company hopes to gain new insight on human behavior by watching what its users do in aggregate. One day, for example, a company might use Aardvark’s technology to help people communicate quickly across large organizations. A multinational corporation with tens of thousands of users could use it so a salesperson in one country could instantly ask a question of the right technologist in another country — more efficient than an email list. To that end, the company is already testing out a way to create groups of users within the site around particular themes.
Which brings us back to mobile. One of the most obvious use cases for wanting to get an expert answer from someone you know is when you’re away from the computer — like when traveling in a foreign city and trying to find a fun place to go out at night. You can already use Aardvark on your phone using mobile email or a third-party mobile IM client, but the company aims to make it even easier through mobile applications and SMS. First up for development is an iPhone application, Keighran (pictured) says, that will let you either type in a question or speak it into the phone. The message will then go into the Aardvark system, appearing in front of the experts you’re connected to on the site — sort of like a local review site, but live and just with your friends.
Look for the iPhone app in the next two months.
[Photo of Keighran via The Guardian.]
VB's research team is studying web-personalization... Chime in here, and we’ll share the results.