In this paper are considered questions which connected with building of the hybrid user model in scientific recommender systems. Offered model is consistent with conceptual model of the presentation scientific CERIF data (Common European Research Information Format). The authors propose an approach using semantic networks. The model considers the main entities of a physical data model, received from generally available sources. During creation of model approaches from area of marketing were used, in particular psychographic approach. Quantitative experiment for profiles of the users of the “International congress conference «Information technologies in education»” portal. Results of experiment have confirmed adequacy of the offered models. The offered detail level of model and methods of its representation are directed to further improving of a pertinent of all personalisation algorithms. This study was financially supported according to Federal Target Program “Research and development on priority directions of scientific-technological complex of Russia for 2014-2020” (grant № RFMEFI57614X0068).