Open-domain Commonsense Reasoning Using Discourse Relations from a Corpus of Weblog Stories
Matthew S. Gerber, Andrew S. Gordon, Kenji Sagae · 2010
We present a method of extracting opendomain commonsense knowledge by applying discourse parsing to a large corpus of personal stories written by Internet authors. We demonstrate the use of a linear-time, joint syntax/discourse dependency parser for this purpose, and we show how the extracted discourse relations can be used to generate opendomain textual inferences. Our evaluations of the discourse parser and inference models show some success, but also identify a number of interesting directions for future work. 1