CMUQ@QALB-2014: An SMT-based System for Automatic Arabic Error Correction

Abstract

In this paper, we describe the CMUQ system we submitted to The ANLP-QALB 2014 Shared Task on Automatic Text Correction for Arabic. Our system combines rule-based linguistic techniques with statistical language modeling techniques and machine translation-based methods. Our system outperforms the baseline and reaches an F-score of 65.42% on the test set of QALB corpus. This ranks us 3rd in the competition.