These recommendations are generated automatically from the titles and descriptions of the resources. A multilingual language model analyzes the content and identifies which resources are thematically similar. Based on this, additional relevant resources related to the currently selected one are displayed, even if different terms are used.