Towards Logical Inference for Arabic Question-Answering

This article constitutes an opening to think of the modeling and the analysis of Arabic texts within a question-answering system. It is a question of exceeding the traditional investigations focused on morpho-syntactic approaches. We present a new approach that analyzes a text, transforms it to logical predicates and extracts the accurate answer. In addition, we represent different levels of information within a text and choose an answer among several proposed. To do so, we transform the question and the text into logical forms. Then, recognize all entailments between them. So, the results of this recognizing are a set of text sentences that can implicate the user’s question. Now, our work is concentrated on an implementation step to develop a question-answering system in Arabic using the techniques of textual entailment recognition. Text features extraction (keywords, named entities, relationships that link them) is actually considered the first step in our text modeling process. The second one is the use of textual entailment techniques that relies on inference and logic representation to extract the candidate answer. The last step is the extraction and selection of this answer.