Commonsense Knowledge Types Identification andReasoning for the Winograd Schema Challenge


Nov 17, 2018 AKBC 2019 Conference Blind Submission readers: everyone Show Bibtex
  • Keywords: Commonsense Knowledge Types Identification, Winograd Schema Challenge, Commonsense Reasoning
  • TL;DR: In this paper we describe various new commonsense knowledge types by exploiting a Winograd Schema Challenge dataset which will become helpful in crowd-sourcing or automating knowledge extraction.
  • Abstract: The Winograd Schema Challenge was proposed as an alternative to the Turing test in 2011. It is a text based question answering challenge which, according to its creators,“requires world knowledge and default reasoning abilities”. In this work, we performed a comprehensive study of the problems in the challenge from the perspective of identifying the types of knowledge which are needed to solve them. We identified 12 knowledge types to address the entire challenge corpus. We also defined a logical reasoning algorithm to tackle 10 out of 12 knowledge types. The algorithm covers 82.47% of all the problems in the dataset. Furthermore, we show how the overall approach generalizes on a well known pronoun resolution corpus which is inspired from the Winograd Schema Challenge.
  • Archival status: Non-Archival
  • Subject areas: Question Answering, Reasoning, Knowledge Representation
0 Replies