Abstract
Language Tutoring Systems (LTSs) are computer-based tutoring systems that address linguistic skills. Most LTSs rely on Natural Language Processing (NLP) tools for analyzing the student's answer to support him/her with appropriate feedback. In such systems, diagnosing the student errors is typically based on a large library of students' expected bugs. In addition, such systems (especially in case of the Arabic language) don't consider modeling of the student knowledge or assigning difficulty levels to the presented questions. Subsequently, such systems do not support individualization of their interactions to each student's needs and preferences. This paper presents the Arabic Grammar Tutor (AGTutor), which utilizes Constraint-Based Modeling (CBM) to model the domain eliminating the need for the bugs' library. Grammar rules are represented as a set of constraints that are categorized in terms of the different linguistic skills. Linguistic skills are classified into several levels and categories. Such hierarchical structure allows diagnosing the student errors in a more efficient way. In addition, modeling the student competence in the Arabic linguistic skills is also expressed in terms of a set of domain constraints. Assigning difficulty levels to each question according to the required skills and modeling the student knowledge allow the system to adapt its interactions according to the student week points to enhance his/her Arabic linguistic skills. Evaluation results have shown the effectiveness of AGTutor in tutoring the students and the satisfaction of both the students and the teachers with its features and capabilities.