Josip Jukić is a researcher at TakeLab, FER, pursuing a Ph.D. in active learning — a family of machine learning algorithms designed to reduce the cost of developing machine learning models.
He specializes in natural language processing and has several research interests, such as the interpretability of deep learning models and automated text-based personality assessment. Josip is the cocreator of ALANNO, an open-source annotation system for natural language tasks powered by active learning.
ALANNO: An Active Learning Annotation System for Mortals
In today’s data-driven society, supervised machine learning is rapidly evolving, and the need for labeled data is increasing. However, obtaining labels is often an expensive and tedious process that requires human effort. To address this problem, we developed ALANNO, an open-source annotation system for natural language tasks powered by active learning.
We have designed solutions that make active learning effective in real-world applications to achieve impressive model performance with minimal annotation overhead. We support the system with a wealth of state-of-the-art active learning methods, including Bayesian approaches and gradient-based strategies specialized for deep learning models.