Recommender Systems Based on Detection Community in Academic Social Network

Boussaadi*, Smail, Aliane**, Hasina, Abdeldjalil Ouahabi***
2021 Zenodo  
Academic social networks have become essential online platforms for researchers seeking digital visibility and scientific collaboration. But, the speed with which new scientific articles are published and shared on these academic ecosystems generates a situation of cognitive overload, creating disorientation in the search for relevant and useful information, hence the need for a filtering process to reduce this information overload. In this context, we propose a hybrid approach to recommending
more » ... cientific papers that uses collaborative filtering combined with semantic exploration and extraction of latent themes by techniques combining generative and inferential aspects such as the Dirichlet latent allocation (LDA) technique. A key element of our approach is the representation of a researcher's profile by topics closely related to his field of interest. We model the interactions between researchers by a weighted graph so that researchers sharing the same thematic interests will be grouped into meaningful communities. The Jensen-Shannon Divergence (JSD) is used to determine the extent to which a scientific article is similar to a researcher's interest profile. The integration in our approach of a community concept based on latent topics reduces the problem of data sparsity since the recommendation process is limited to a small amount of space, thus reducing processing time and data storage space (researcher-articles). Experiences confirm our performance expectations by showing good relevance and objectivity in the results
doi:10.5281/zenodo.5801852 fatcat:wufjafk54bhe5o25f5zrvonqyi