Inspiration

Dating apps like Tinder focus solely on the appearance of the users. But there is a lot more to dating than just looks. One of the most important thing that determines compatibility is humor! "I always have a good time with her/him because they make me laugh!" Long lasting relationships require more than just acceptance of physical looks and humor is one of top most priorities most people look at while trying to find a partner.

Now find a partner, doing what you normally spend time on when using social networks: reading jokes and memes!

What it does

We found a unique way to quantify humor. The app presents a series of jokes to you. You must respond to it with a 'Hah!' or a 'Nah..' and each response is recorded to create your humor profile. Once a match occurs, the matched are given one chance to speak to each other over the phone (using hyphenate) for 30 seconds post which the matched must decide if they want to move ahead. If they do, they can now chat inside the app (also using hyphenate)! Using our advanced matching algorithm, that classifies joke using their natural language properties and their relative likelihood of being liked/disliked across the entire user graph, we are able to match similar humor profiles in real time! Find you LoLmate!

How we built it

Our team of two began working with one team member designing the frontend app (iOS) while the other took to designing the backend and implementing the matching algorithm. The matching algorithm took quite some time and has taken inspiration from the eigenvector method of Google's page ranking to determine joke ranks which enable us to infer about jokes which haven't been seen by a user yet.

Challenges we ran into

Data was too sparse and coming up with an algorithm to handle such data was difficult. The frontend design was a challenge because presenting jokes and capturing response without having any distraction of being in a dating app was an impediment.

Accomplishments that we're proud of

We've implemented the algorithm and it is able to make really good predictions (on comparing the vectors of humor profiles for individuals matched using dummy data). We're also very proud of our design! Minimalistic and straight to the point.

What we learned

Data problems are far harder than what they look like.

What's next for LoLMate

Improve the algorithm and add more features like the ability to create content.

Built With

Share this project:
×

Updates