Representing Knowledge about Norms to Reason on Texts

Norms are essential to extend inference: inferences based on norms are far richer than those based on logical implica-tions. In the recent decades, much effort has been devoted to rea-son on a domain, once its norms are represented. How to extract and express those norms has received far less attention. Extraction is difficult: as the readers are supposed to know them, the norms of a domain are seldom made explicit. For one thing, extracting norms requires a language to represent them, and this is the topic of this paper. We apply this language to represent norms in the do-main of driving, and show that it is adequate to reason on the causes of accidents, as described by car-crash reports.