7/21/2016 By Mike Koon, College of Engineering
Written by By Mike Koon, College of Engineering
A group of four CS @ ILLINOIS students claimed top prize at HackGT2014, a hackathon held at Georgia Tech and sponsored by Tech Square Labs. The group of Kevin Jasieniecki, Keagan McClelland, Kevin Scheer and Nathan Dolph was one of three teams from Illinois that finished among the top 10 in a contest that attracted about 1,000 college students.
A growing phenomenon that has its roots in corporate America, college hackathons are typically held on a campus and give participants 36 hours to code an app. Although each team can come to the contest with an idea, they must not bring any previous code with them.
For coding an app called PickMeUp, the foursome won $10,000 in cash, $60,000 in Microsoft server credits, $50,000 in investment funding from TechSquare Labs, a provisional and utility patent, and a second round of interviews with Trip Advisor.
The idea behind the app is an interesting take on natural language processing (NLP), which uses the connection of language to determine mood and thus predict behavior. In this particular case, it would take a series of text messages from a phone to determine the mood of the user and if he is sad, automatically send a car to their home and prompt them to go out on the town.
“We decided to use the final hours to figure out how we could extend this concept and make it more applicable to the real world and practice our pitching skills,” McClelland said. “That ended up catching a lot of the judges’ attention.”
“It’s ‘a Pick Me Up that picks you up,’” Jasieniecki said of the ultimate tag line. “When we said that line at the end of the pitch, it really struck a chord with audience and ultimately the judges.”The team actually scrapped its initial idea eight hours into the contest figuring the concept of taking legal text and abbreviating it into simpler language would take much longer than the 36 hours to create. Once settling on the eventual winning idea, the team fed 400,000 tweets through Python SKLearn in order to train a naïve Bayes classifier. The goal was to determine whether the tweets were either negative or positive based on the emoticon.
“Once it reached a certain threshold of accuracy based on the conditional probability, we applied it,” Jasieniecki said. “This is a very hard NLP problem of which we have only scratched the surface, but it made for some fun experimentation.”
“Although the judges indicated the use case might seem limited, they felt we had stumbled on an idea that will do very well given a different application,” McClelland said.
For instance, Jasieniecki wrote in his report, “Let’s just say there is a café that paid for live music, and your café is empty. The coffee shop now has an easy way of tapping into who might be having a bad day and could give a deal to those customers who were accessed with this service.”
For Jasieniecki, the win is culmination of success in his hackathon experience. He has 10 stickers on his laptop, one for each event he has participated, and has received each of his internships through contacts from previous hackathons. His team finished in the top five in a contest at Virginia Tech in the spring and reached the second round in judging at Michigan this fall, but this is the first time he has been on a team that has won top prize.
In the next few weeks, the group will be setting up a business account to house the investments and fine-tune the process for more reiterations.
“This is where more serious rubber meets the road as we take the specs and prepare for an initial build with TechSquare,” McClelland said.