A Probabilistic Setting and Lexical Cooccurrence Modelfor Textual Entailment

摘要

This paper proposes a general probabilisticsetting that formalizes a probabilisticnotion of textual entailment. We furtherdescribe a particular preliminary modelfor lexical-level entailment, based ondocument cooccurrence probabilities,which follows the general setting. Themodel was evaluated on two applicationindependent datasets, suggesting the relevanceof such probabilistic approaches forentailment modeling.