Abstract:

Determining the textual entailment between texts is important in many NLP tasks, such as, question-answering, summarization, and information retrieval, etc. In question answering, this technique is frequently used to validate an answer retrieved by a question answering system. In this paper, we address the problem of textual entailment in Arabic. We employ some features to determine the textual entailment between pairs of logical representations of a question and a passage of text. We have implemented our approach in an Arabic recognizing textual entailment system called Ar-SLoTE (Arabic Semantic Logic based Textual Entailment). The experiments results of the entailments classification achieved a precision and a recall successively equal to 73% and 68% on average.