By Robin Brooks
Four of the 2017 eScience Institute’s Data Science for Social Good (DSSG) fellows took first place in the Seattle Global AI Hackathon, “the world’s first ever AI hackathon series”. Brett Bejcek, Mayuree Binjolkar, Orysya Stus, and Michael Vlah met during the DSSG program and designed a fake news detector called Breaking News based on a Naive Bayes Text Classifier.
The application analyzes the text of a reader-supplied article (not the headline or news sources) and determines what percentage of it is “fake” news. It also offers a sentiment analysis of the article and the similarity of the body and heading texts.
They estimate that the program is about 85% accurate, and are considering increasing the number of features in the program and the complexity of their model. In fact, through good communication and effective teamwork, they nearly had a fake image analysis function, but they ran out of time.
Despite the short amount of time they’ve known each other (three weeks) their work within the DSSG program and the hackathon has made them “best friends”, and the four team members often expressed appreciation for each others’ skill sets, such as programming, back-end development and problem-solving. “It’s far too dangerous to dream too small than too big. It was really cool to see everyone work together,” said Bejcek.
The Breaking News application will now go on to compete against the winners from the other 14 cities who participated in the hackathon; over 4000 people competed globally in the event, which happened June 23 – 25. The students have ambitions to continue working with one another, and are planning to attend two more hackathons this summer.