The LexTime Dataset

LexTime is a dataset built to measure the ability of LMs to perform temporal event ordering on a specific type of language: legal language. Each instance contains: a context paragraph taken from a complaint document a pair of events, which can be either one implicit event and one explicit event or two explicit events. An event is an occurrence or action triggered by a verb or noun that takes place at a specific time and can be ordered relative to other events based on temporal relationships (TR) a query consisting of a pair of events and a temporal relationship between them, creating a triplet following this template: {event A, TR, event B}