Junebug online dating site hunts users to test their game-changing algorithms

Online dating sites ought to have nearer the success rate as a search engine, says John Myles White.
“When you compare the people that big dating sites suggest to you with the pages that Google gives you in response to a search, the difference is staggering,” says Myles White, a Ph.D candidate in Princeton’s psychology department.

Last week, Myles, 28, and partner Jim Keller, 29, the founder and CEO of Willow Grove-based web development and strategy company Context, announced the launch of Junebug, what they call their answer to “the lack of innovation in online dating.”
The duo is entering the crowded online dating scene because they say their competition isn’t leveraging contemporary statistical techniques to their fullest extent. Now all they need are the users and data to prove it.

So on July 4, they threw Junebug open to the public, focusing on bringing on users from the crowded Philadelphia and New York City markets. With enough users ranking other profiles, in time they say their algorithms will make Junebug among the most successful online matchmaking services out.

“I’ve become increasingly convinced that the big players in the online dating market are either unaware of or simply indifferent to the tools that computer scientists can offer them,” says Myles White, a native of Hoboken, N.J..
Keller’s hand comes in developing Junebug with a level of user-friendliness that has become expected throughout the web but that the pair says most online dating sites haven’t met.

The pair says a short-term goal is to establish advertising as a revenue model for the self-funded project, but the focus is on bringing on users to make sure their product is everything they hope it to be. Which is a good goal because they’re making a big swing — not only railing against free dating competitors, sites like PlentyoFish and OKCupid, but also the paid giants like eHarmony and Match.com, though Keller says he doesn’t see free sites being in direct competition with paid versions.
“Perhaps the paid sites are doing something clever, but it’s hard to believe that given the experiences that anyone I’ve known has had with those sites,” says Keller who is charged with the heavy development of the site. “Reading through their patents suggests that they’ve played around with things like neural nets, but never invested much work into using more contemporary techniques like SVMs, Bayes Nets or NMFs.”
Deep math aside, the pair seems to be saying they can do to online dating what sabermetricians have done to baseball. That’s not to say, the web isn’t nearly a decade deep into concerted efforts to harvest data to improve online matchmaking. Of late, prominent movement has been made to focus on data elsewhere online. Fellow free service OKCupid has created something of a following around its popular blog known for crunching the data it collects.
But that, too, Keller says, shows a weakness.
“[The OKCupid blog] suggests that they’ve not begun working with developers who have a real mastery of modern statistical techniques. I’ve been surprised by how many of their analyses … are conducted using the most basic statistical techniques,” says Keller, a native of the Northeast who now lives in Hatboro.
Keller says that contrasts with what he and Myles White are doing.
“Junebug is built on a totally automated backend that only needs normal end-user inputs to provide the data for our algorithms. We’re asking users to rate the profiles they read on a scale from one to 100, which, amazingly enough, seems to be something that other dating sites have never pushed, even though it’s precisely explicit rating that makes the algorithms that won the Netflix prize so powerful,” Keller says. “In fact, those algorithms were so fascinating because they were able to extract semantic information about genres and other structure in the data set that were implicit in the ratings.”
Keller says they’re using a similar set of algorithms, in addition to other approaches like automated text and image analysis.
Though user acquisition is the next step, Keller speaks surely: “I’m confident that these tools have not been used by existing free dating sites and that they can provide substantially better matches than existing dating sites.”
-30-
Every Monday, Technically Not Tech will feature people, projects, and businesses that are involved with Philly’s tech scene, but aren’t necessarily technology focused. See others here.