An Integrated Method To Recognize Textual Entailment

Textual entailment recognition is an important research area of natural language processing and information extraction. In order to solve the problem of variability of semantic expressions in natural language processing and slot filler validation, this paper studies and analyzes the technology of textual entailment recognition. The slot filler validation track focuses on the refinement of output from English slot filling systems by either combining information from multiple slot filling systems, or applying more intensive linguistic processing to validate individual candidate slot fillers. In addition, textual entailment recognition is useful in question answering, summarization, information extraction, semantic searching, and machine translation.This paper proposes an integrated method of textual entailment recognition, and develops a system of textual entailment recognition. Meanwhile, this paper proposes two kinds of classification features which include features based on semantic roles, and ones based on dependency relations and WordNet. Firstly, this paper utilizes two approaches to identify the entailment relation between texts and hypothesis. The first approach is based on lexical similarities whilch utilizes five kinds of lexical methods based on cosine similarity, longest common subsequence, edit distance, common words and skip n grams. The second approach is based on the classifier of support vector machine which means to treat this task as a classification problem. The features of the classifier include lexical features, syntactic features, lexical sementic features and syntactic sementic features. Secondly, the proposed integrated method in this paper improve the effect of the recognize system. The experient result of the integrated approach is better than the results of the approach based on lexical similarities and the approach based on SVM. Hence, the integrated approach is suitable for the textual entailment recognition.