Multi-label models for Java, Python, and Pharo are trained and evaluated on 10,555, 2,555 and 1,765 class-level code comments, respectively, extracted from 20 open source projects. The results ...