Arabic Pronunciation Assessment for Saudi Arabian Students: Corpus Development and System Architecture

The reading skill constitutes a principal pillar of education. The existence of any deficiency in reading may negatively affect a student’s progress in other subject areas. This research study primarily aims at improving the Arabic reading skills of primary school students in Saudi Arabia. Specifically, it lays the groundwork for developing an interactive Arabic reading tutor using a machine-learning/deep-learning web-based platform. This platform assesses students’ reading of specific Arabic paragraphs and provides instant feedback. The study encompasses the development of a speech corpus tailored to the age range of 6–15 years. The speech corpus has been developed with over 503 participants and resulting in 33 h of audio corpus, making it the largest speech corpus designed for pronunciation assessment in Saudi Arabia. Within this system, student readings are recorded and saved in a database, allowing Arabic teachers to manually assess the reading and edit any incorrect automatic feedback, if necessary, according to the rules that govern the proper recital of modern standard Arabic. The platform will aid teachers in identifying issues and mistakes in the readings, actively involving them in improving the performance of the intelligent assessment tool. The proposed system architecture leverages technologies to provide targeted and effective feedback on pronunciation, ultimately contributing to an enriched language acquisition process.