On March 8, 2019, machine reading finally got rid of the “spell”.
Cloudwalk and Shanghai Jiao Tong University came up with a brand-new model based on the original DCMN algorithm.
This model topped the list on the ReAdingComprehension (RACE) dataset collected from English examinations, which marked the AI system’s surpassing human beings in deep reading comprehension for the first time.
Address of the paper: https://arxiv.org/abs/1908.11511?context=cs.CL
