QARAB

Abstract

We describe the design and implementation of a question answering (QA) system called QARAB. It is a system that takes natural language questions expressed in the Arabic language and attempts to provide short answers. The system's primary source of knowledge is a collection of Arabic newspaper text extracted from Al-Raya, a newspaper published in Qatar. During the last few years the information retrieval community has attacked this problem for English using standard IR techniques with only mediocre success. We are tackling this problem for Arabic using traditional Information Retrieval (IR) techniques coupled with a sophisticated Natural Language Processing (NLP) approach. To identify the answer, we adopt a keyword matching strategy along with matching simple structures extracted from both the question and the candidate documents selected by the IR system. To achieve this goal, we use an existing tagger to identify proper names and other crucial lexical items and build lexical entries for them on the fly. We also carry out an analysis of Arabic question forms and attempt a better understanding of what kinds of answers users find satisfactory. The paucity of studies of real users has limited results in earlier research.