A concept-based knowledge representation model for semantic entailment inference
Meijing Zhao, Wancheng Ni, Haidong Zhang, Yiping Yang · 2014
Semantic entailment is a fundamental problem in natural language understanding field which has a large number of applications. Knowledge acquisition and knowledge representation are crucial parts in semantic inference strategies. This paper presents a principled approach to semantic entailment problem that builds on a concept-based knowledge representation model (CKR). This model formally defines the concept as a triple (attribute, relation and behavior) and the knowledge of a concept can be illustrated by the triple. We propose a semantic inference strategy that against identify text segments which with dissimilar surface form but share a common meaning. The inference strategy avoids syntactic analysis steps. A preliminary evaluation on the PASCAL text collection is presented. Experimental results show that our concept-based inference strategy is effective and has strong development potential.