Abstract:

Definitional questions and factoid questions are commonly used in recent Arabic Question Answering (QA) works. These questions are simple  questions  which  can be answered  by a simple answer or a Named Entity (NE). Most Arabic QA systems lack the capability of processing complex ques- tions, namely Why-questions. Motivated by this issue, we release a new QA system aiming to answer Arabic Why-questions in the medical field. Therefore, we propose a novel approach using Natural Language Processing (NLP) and In- formation Retrieval (IR) techniques. Our system uses dictionaries and transduc- ers to analyze Why-questions and to extract causal relations which represent the answer of a Why-question. Compared to previous studies, our approach obtains comparable results on processing Why-questions.