Graphics and illustrations by Lauren Sisk.

Editor’s Note: We’re developing one of the most comprehensive interest-based matching systems for a dating site, ever. Geeks are very particular about their interests; for instance, if someone says I like “Final Fantasy” the next automatic question from another Final Fantasy is often “which ones?” or “before or after VII/XI/etc?” Our matching algorithm reflects that level of specificity and goes even deeper.

My first really big project in creating Project DragonFruit was to create a novel and effective matching algorithm.  Fascinating, I know.  For those of you still reading, it actually gets pretty interesting.

DragonFruit Algorithm
Like sorting out the conceptual ideas on how to get our algorithms to break down fandoms

 

IMG_0722
The math-y stuff to ensure that fandoms get broken down and not our computational power instead, not so much. 

The trick in making something like this work properly is not to have a simple 1-1 matching, because that would omit tons of valuable data from the matched set.  For example, if someone is a fan of the Star Wars Extended Universe, obviously the best match will be someone else who is into the Star Wars Extended Universe, but that won’t address anybody who is into the Star Wars movies.  We want to make the matching be as effective as possible with minimum of effort from the user, so we needed to get creative with the way we match. What we came up with was a ginormous tree of possible geekdom entries, each nested beneath more generic geekdom values:

DragonFruit AlgorithmThe benefit of this is that someone searching for simply Star Wars will find SW EU geeks, Original trilogy fans, even the odd Jar-Jar Binks aficionado, whereas a person who specifies Original Trilogy will remain blissfully Binks free.  This incentivizes specificity without compromising on result quality, and even allows for other trees to be exploited in a limited way.  If someone is so boring as to not be able to think of anything more interesting than Sci-Fi, they will find general fans of Star Wars, Star Trek etc, but be saved from the true fanatics who only care about Ocampa lore (from ST:VOY).

Finally, in order to prevent huge drop downs, I created a natural language predictor which will fill in the geekdom of your choice in a format that matches ours, and made sure that if you put in something we missed completely, we can figure out where it fits and get you rolling ASAP.

Got questions or suggestions? We’d love to hear them in the comments.

If you’re from the media (this includes bloggers, podcasters, etc), read our product announcement press release for further details about the app and contact info.

 

Find that geeky significant other in your life.
Join the web beta for the first true mobile dating app for geeks, coming soon for iOS and Android!

 

CREATE MY ACCOUNT
Like what you see and want to share it with your friends/enemies/lovers?
Share on Facebook0Tweet about this on TwitterShare on Google+0Pin on Pinterest0Share on LinkedIn0Share on Tumblr0Email this to someone