Paper-oriented academic recommendations are crucial for guiding users towards relevant papers based on their behaviors and the content of the papers. However, existing recommendation models heavily rely on user-paper interactions, leading to suboptimal performance when encountering novice learners who lack explicit academic behaviors. In response to this challenge, we introduce a novel approach called Social Network-based Academic Recommendation approach (SNAR) to predict the needs of novice learners by leveraging their social connections. Specifically, SNAR establishes connections between novice learners and other researchers through mentor-student relationships, forming a guidance pool of pertinent papers. Utilizing the attentive mechanism on user profiles (e.g., research interests) and paper descriptions (e.g., titles and keywords) based on the guidance pool, SNAR produces precise representations for both users and items. We collect an academic dataset of novice learners and conduct extensive experiments. The results demonstrate the effectiveness of SNAR in providing targeted paper recommendations for novice learners.