Moving QA Towards Reading Comprehension Using Context and Default Reasoning

For a question answering system to understand the underly- ing assumptions and contextual features of natural language, a contextually and semantically sensitive knowledge repre- sentation and reasoning module is essential. This paper pro- poses a three-layered knowledge representation coupled with reasoning mechanisms for defaults and contexts as a solution for pushing question answering towards a state capable of ba- sic reading comprehension. A travel scenario is presented as a vehicle for demonstrating the complexities of this problem, and the ways in which semantic relations, contexts (for con- ditionality, planning, time, and space), and defaults address these complexities.