"What is Learned in Visually Grounded Neural Syntax Acquisition", Noriyuki Kojima, Hadar Averbuch-Elor, Alexander Rush and Yoav Artzi (ACL 2020) - View it on GitHub