Towards a corpus based heuristic for assessing communal common ground

In this article we present a corpus-based heuristic for determining what facts are likely to be known by most or all the members of a given community. We believe that a heuristic of this kind could be useful in a large range of applications, computational and otherwise. For example, the heuristic could help a Natural Language Generation system to avoid stating facts that are widely known, thus diminishing the chance that the reader will get bored or irritated. Another example is a system that uses formal argumentation theory to present arguments to people; such a system could be improved by linking arguments to facts that are widely known. For example, to argue that butter is unhealthy, it might suffice to say “butter increases cholesterol”, because it is widely known (or believed) that cholesterol can cause heart disease. Our goal is to use the proposed heuristic in determining what facts to mention in a description of a famous person.