Mapping Texts to Scripts: An Entailment Study

Mapping Text to Scripts: An Entailment Study Simon Ostermann, Hannah Seitz, Stefan Thater, Manfred Pinkal Saarland University Saarbr¨ucken, Germany {simono|hseitz|stth|pinkal}@coli.uni-saarland.de Abstract Commonsense knowledge as provided by scripts is crucially relevant for text understanding systems, providing a basis for commonsense inference. This paper considers a relevant subtask of script-based text understanding, the task of mapping event mentions in a text to script events. We focus on script representations where events are associated with paraphrase sets, i.e. sets of crowdsourced event descriptions. We provide a detailed annotation of event mention/description pairs with textual entailment types. We demonstrate that representing events in terms of paraphrase sets can massively improve the performance of text-to-script mapping systems. However, for a residual substantial fraction of cases, deeper inference is still required. Keywords: script knowledge, annotation study, textual entailment 1. Introduction and Motivation Scripts represent knowledge about everyday activities, or scenarios, like “going to the movies” or “having dinner at a restaurant” (Schank and Abelson, 1975). They consist of events like ORDER MENU or EAT that take place in the sce- nario, plus information about the typical temporal order in which these events happen, as well as knowledge about par- ticipants that play a role in the script, such as the waiter or plates and cutlery. Figure 1 shows a graphical representation of the script of the BAKING A CAKE scenario. The nodes in the graph represent events and the edges their temporal order. Relevant script participants, such as CAKE, KITCHEN or INGREDIENTS, are listed in the lower right corner. The dashed boxes represent possible linguistic realizations of the individual events in terms of paraphrase sets, described in more detail below. MIX INGRE- DIENTS BAKE WAIT - mix ING - stir ING together - stir well - … - put C into O - bake C - … - wait for a while - wait - … CHOOSE RECIPE BUY INGRE- DIENTS GET INGRE- DIENTS - get ING from S - buy ING - go to S to buy ING - … - get ING for C - collect ING in K - … - browse for R - look for a R - select R - … ING - S - R - C - K - O - cake mix, ingredients, eggs, sugar, … shop, mall, … recipe, cookbook, … cake, pastry, … kitchen, … oven, … Figure 1: An example for a temporal script graph (BAKING A CAKE). In communication, script knowledge is assumed to be part of the common ground, and people often do not mention events which can easily be inferred to have happened by the addressee. For instance, if someone tells about the last time they baked a cake, it is likely that they do not mention the fact that the cake was put into the oven, because it is obvious that this event took place. In contrast, a text understanding system that does not have access to script knowledge will probably not be able to draw this inference. MIX INGRE- …

Content truncated to 500 words