Acquiring temporal constraints between relations

We consider the problem of automatically acquiring knowl-edge about the typical temporal orderings among relations (e.g., actedIn(person, film) typically occurs before wonPrize (film, award)), given only a database of known facts (rela-tion instances) without time information, and a large doc-ument collection. Our approach is based on the conjecture that the narrative order of verb mentions within documents correlates with the temporal order of the relations they rep-resent. We propose a family of algorithms based on this con-jecture, utilizing a corpus of 890m dependency parsed sen-tences to obtain verbs that represent relations of interest, and utilizing Wikipedia documents to gather statistics on narrative order of verb mentions. Our proposed algorithm, GraphOrder, is a novel and scalable graph-based label prop-agation algorithm that takes transitivity of temporal order into account, as well as these statistics on narrative order of verb mentions. This algorithm achieves as high as 38.4 % ab-solute improvement in F1 over a random baseline. Finally, we demonstrate the utility of this learned general knowledge about typical temporal orderings among relations, by show-ing that these temporal constraints can be successfully used by a joint inference framework to assign specific temporal scopes to individual facts