Description:"We explore interpretability in the context of
QA on a combination of KB and text. In particular, we apply attention, LIME and input
perturbation (IP)."