Students of FER's Text Analysis and Knowledge Engineering Lab (TakeLab) participated again this year at the SemEval competition on natural language processing. At SemEval-2017, students competed in four challenging tasks dealing with the application of machine learning in computational semantic analysis of texts. This time around, TakeLab's students ranked 2nd on the task of humor detection in tweets and 3rd among 16 teams on the task of tweet sentiment detection.
Under a watchful eye of TakeLab's members, students Marin Kukovačec, Toni Kukurin, David Lozić, Juraj Malenica, Ivan Mršić, Lukrecija Puljić, Filip Šaina, Antonio Šajatović, Doria Šarić, and Ivan Tokić developed systems capable of detecting sentiment of a tweet given tweet's topic, predicting which tweet from a pair of tweets is more humorous, and finding topically similar questions and best answers on community Q&A platforms such as StackOverflow.
Each year, SemEval competition attracts numerous teams from the best universities around the world. This time around, TakeLab's students ranked 2nd on the task of humor detection in tweets and 3rd among 16 teams on the task of tweet sentiment detection. The results are even more impressive considering that all participating students were bachelor students, who haven't yet had a chance to enroll machine learning and text analysis courses.