This project will bring our top-level academic research into contact with real-world applications of DL. Our strategic and long-term collaboration between academic and non-academic partners will benefit our industries and our society and will advance the science of DL and interpretability itself. Understanding which classes of applications require which DL analysis techniques is the core of our proposed project. In this project, we investigate various methods for interpreting DL-based learning systems. This will be done in a coherent group of interrelated complementary work packages operating in three different domains (text, speech, and music) and in a way that bridges academic and commercial partners, the expert user and the non-expert end user, and various theoretical and practical approaches.