Article Preview
Top1. Introduction
In the modern world, diffuse or ubiquitous computing enables the interconnection of all areas of life using new technologies to foster the creation of intelligent environments, such as smart campuses. It provides reliable Internet-of-Things (IoT) services to the users anytime and anywhere. Recently, the paradigm of smart learning has become popular because of its huge coverage of topics from around the world. It plays a vital role for learners to open the door to new opportunities on higher education. It introduces learning materials in a style that enhances learners' expectations and interactions with each other using different multimedia contents, such as graphics, audio, video, and text. However, a large number of resources is shared in an intelligent learning environment every day causing an overload of information exchanged. This can make it difficult to extract better content. Nevertheless, Arabic language is rich of morph-syntactic and semantic features that complicate its analysis (Mahmoud and Zrigui 2019a). It represents a fundamental problem in Natural Language Processing (NLP) due to the wide variety of applications associated with it (e.g., information retrieval, question-answer, temporal information retrieval, etc.) (Haffar et al., 2020a; Haffar et al., 2020b). Faced with these problems, Arabic documents recommendation-based approach is proposed. The main objective is to find the most suitable content that meets the needs of Arabic users. It facilitates the collaboration between users and improves their learning experiences in different situations. To tackle the sparsity and unbalanced ratings distribution problems, the main contributions of this work are the following:
- •
The use of users generated ratings and the title of Arabic documents as inputs.
- •
The enrich of latent features from items by using the advantages of deep neural networks.
- •
To the best of our knowledge, it is the first work focusing on Arabic resources recommendation within smart campus exploring the relative importance of heterogenous data sources (ratings and document title) for rating prediction task. Moreover, the complexities of Arabic language are processed using deep learning architectures and NLP techniques.
This paper is structured as follows: First, we present the general context of recommendation systems in smart campus. Then, a state of the art on mobile learning field reviews the existing personalized recommendation-based methods. Subsequently, the problems concluded are summarized. After that, we detail the motivation and the phases constituting our proposed approach. Finally, the last section describes a conclusion and our future works.