Learning to Research: Learning to Ranking the Similar Papers via BERT Fine-Tuning

Abstract

Information retrieval has always been an important research topic in the era of big data. How to accurately retrieve the references needed by scholars from a large number of papers, sort them by relevance, and screen out valuable information to recommend to scholars is an important research demand at present. There has not yet been a model that can capture user attention in academic scenarios based on search results. At the same time, the construction resources of this dataset are limited, and there is a lack of a benchmark dataset that can be used for training search learning models. Therefore, this article constructs a dataset that can be used for academic literature relevance ranking search and a learnable search ranking model. The test results indicate that the model has different advantages in disciplinary fields, confirming that this system can be well applied in academic scenarios and improving the efficiency of scholars in literature search and selection.