Annotating Lexically Entailed Subevents for Textual Inference Tasks

摘要

This paper presents a procedure for constructing an Event Structure Lexicon (ESL), a resource which represents the lexically-entailed subevents in text as a support for textual inference tasks. The ESL is used as a resource for a subevent markup algorithm, called subevita, which annotates event implicatures on top of TimeML-based extraction algorithms. Such a resource can be used independently within the RTE task and other linguistic reasoning applications. Finally, we present experimental results of the classification for building the ESL of motion verbs in English.