Arabic Question Answering Systems: Gap Analysis

Mariam M Biltawi,Arafat Awajan,Sara Tedmori

doi:10.1109/access.2021.3074950

Abstract

Question-answering (QA) systems aim to provide answers for given questions. The answers can be extracted or generated from either unstructured or structured text. Therefore, QA is considered an important field that can be used to evaluate machine text understanding. Arabic is a challenging language for many reasons; although it is spoken by more than 330 million native speakers, research on this language is limited. A few QA systems created for Arabic text are available. They were created to experiment on small datasets, some of which are unavailable. The research on QA systems can be expanded into different components of QA systems, such as question analysis, information retrieval, and answer extraction. The objective of this research is to analyze the QA systems created for Arabic text by reviewing, categorizing, and analyzing the gaps by providing advice to those who would like to work in this field. Six benchmark datasets are available for testing and evaluating Arabic QA systems, and 26 selected Arabic QA systems are analyzed and discussed in this research.

Highlights

Question answering (QA) is a benchmark task with significant applications for users
The second observation is that before 2015, the document retrieval component was focused on using rulebased techniques, but after 2015, the focus was changed to using either search engines or search techniques because most of the research papers were based on either Web search or document search to retrieve relevant documents
This paper provided a brief introduction on the Arabic language, natural language processing (NLP), and the challenges associated with the Arabic language

Summary

Introduction

Question answering (QA) is a benchmark task with significant applications for users. QA systems aim to provide an answer for a given question extracted or generated from either unstructured or structured text. Community QA systems, generating question systems, and dialog systems are examples of QA systems. While a general QA system aims to make the machine answer questions, other applications focus on other purposes. A community QA system focuses more on information retrieval rather than on answer extraction [1]. A community QA dataset can be created by collecting questions and answers from forums or websites. A dialog system, on the contrary, aims to respond to any type of text by generating a reply in accordance with the given input [3], [4]

Objectives

Methods

Results

Conclusion