Jennifer Chu-Carroll

We propose that the first step in defining MRC tasks should be specifying what content a system would likely need to understand for a given class of applica- ar X iv . Expand

We introduce GLUCOSE, a large-scale dataset of implicit commonsense causal knowledge, encoded as causal mini-theories about the world, each grounded in a narrative context. Expand

We show that a purely generative encoder-decoder architecture handily beats the previous state of the art in FrameNet 1.7 parsing, and that a mixed decoding multi-task approach achieves even better performance. Expand