Application of Machine Learning Models for Recommending a List of Academic Sources

This paper is focused on developing a system to automate the process of searching and selecting academic sources using machine learning methods. It is no secret that the volume of scientific publications in the modern scientific world is growing rapidly, creating significant difficulties for researchers in finding and selecting sources, relevant to their academic work. Thanks to the application of the new machine learning methods for natural language processing, it is now possible to significantly speed up and simplify this process, giving researchers more resources for their work. The primary goal of the system being developed is to provide the user with a list of relevant scientific sources based on the given title and a brief description of their work. This work is based on data from the open arXiv dataset, supplemented with additional information from Semantic Scholar. BERT, CatBoost, and Word2Vec models were applied in this work, each of which was customized to some extent for the given task. Applying this model allows to effectively analyze and process large volumes of scientific papers, ensuring the relevance of the recommendation generated. The developed system will significantly simplify the process of searching for scientific literature, improving the quality of conducted research and saving researchers' time.