Objective At present, literatures are the most important knowledge source of scientific resources. However, in the face of the information overload caused by the explosive growth of scientific literatures, it is difficult for researchers to find the literatures quickly. This is particularly acute in the field of cognitive science. In order to solve this problem, this paper proposes a personalized literature recommendation method based on the domain related interest model. Methods This method is based on the observation about the researcher's short-term interest migration and adopts a time-forgotten curve for modeling user interest. In order to solve the problem of data sparsity, this paper improves the spread activation model to find potential domain semantic relations among interests. Results We use the Pub Med literature as our experimental data. From the point of view of accuracy, f MRI, induction and normal can acquire 0.60, 0.80 and 0.550, respectively. From the point of view of average accuracy, this method can provide higher accuracy and recall rate for the concept of induction. Conclusions The proposed method can effectively capture user interests and recommend literature that more relevant to users' research interests.