Abstract:

The advancement of technologieshas been opened the door to new opportunities on higher education. Recently, mobile learning has been offered greater flexibility in when and where learning happens and enhanced learners' expectations and interactions with each other. In this paper, learning resources recommendation problem is adressed. It allows users (e.g. students or teachers) to discover new educational resources that matches their preferences and continue their learning processes anytime and anywhere. Instead, Arabic learning resources represented an issue due to their richness of features and analysis ambiguities. To deal with their processing complexities, neural-networks based approach is proposed for Arabic learning resources recommendation. The objective was to infer the format of the suitable Arabic document (e.g. text, image, video, etc.) in a contextual situation. Indeed, Convolutional Neural Network (CNN) was efficient to model the interaction between user and document embeddings, create latent representations and predict the top-N recommendations. Given the contextual sensor data (e.g. luminosity, mobility, noise, etc.) from a user, the suitable document with the best format is returned. Experiments demonstrated that this method was useful to adapt interests and context information of users, improve their satisfactions and the quality of learning.