Combining Lexical Resources with Fuzzy Set Theory for Recognizing Textual Entailment

Textual Entailment (TE) recognition is a task which consists in recognizing if a textual expression, the text T, entails another expression, the hypothesis H. Recently it is treated as a common solution for modeling language variability. Textual entailment captures a broad range of semantic oriented inferences needed for many Natural Language Processing (NLP) applications, like Information Retrieval (IR), Question Answering (QA), Information Extraction (IE), text summarization and Machine Translation (MT). Recognizing Textual Entailment (RTE) as one of the fundamental problems in those natural language processing applications has attracted increasing attention in recent years. This paper proposes a new method for textual entailment measure which is based on lexical, shallow syntactic analysis combined with fuzzy set theory. Further we model lexical and semantic features based on this method and perform textual entailment recognition using machine learning algorithm. The performance of our method on RTE challenge data resulted in an accuracy of 56%.