A Contextual Bayesian User Experience Model for Scholarly Recommender Systems

Since the advent of scholarly recommender systems (SRSs), more than 200 papers in the related area have been published. Many of these papers focus on proposing new and more accurate algorithms, or to enhance existing ones. Recently we have seen growing interest in embedding recommending methods into User Experience (UX), to enhance the value of RSs for users. Researchers have proposed that UX can be affected by bottlenecks in human perception, the preconceptions of the individual, and related factors such as personal and situational characteristics, which can be considered as contextual information. Although there are a few studies on developing User Models (UMs) in the field of SRSs, it has been emphasized that incorporating contextual information into user modelling and creating recommendations based on the users’ information needs can be an effective approach to personalization and better UX with SRSs. The aim of this paper is to operationalize relevant contexts and to design a Bayesian UM for assisting the diagnosis of scholars’ information needs in terms of accurate, novel, diverse, and popular research papers. The proposed user model can be embedded in the process of recommending and identifying the users’ information needs which help recommenders to retrieve more appropriate recommendations and consequently leads to the enhancement of the UX for SRSs. Finally, the robustness and performance of the proposed Bayesian UM are evaluated.