Revue Maghrébine des Langues
Volume 9, Numéro 1, Pages 113-130
2014-12-31

High-performance Arabic Question Answering System

Authors : Hamadene Abdelbaki .

Abstract

Question Answering systems are emerged to provide unique, brief and concise answers instead of links toward web pages returned by search engines. Many systems have been deeply thought and efficiently developed for other natural languages especially English. Research studies to develop efficient Arabic Question Answering Systems are still few and need important care to many aspects and Arabic particularities (short vowels, absence of capital letters, complex morphology, etc.). The proposed Question Answering system (ArQA) with sophisticated architecture deals with Arabic factoid questions. It employs the latest approaches of NLP, IR and question answering systems. The experimental results show the effectiveness of our approach based on N-grams matching, question’s focus filtering, and studies about text similarity scoring. Using standard metrics (adopted by TREC), ArQA achieves an overall accuracy of more than 86% and outperforms other systems.

Keywords

High-Performance Arabic Question Answering System