Comparative analysis of keyword extraction strategies

Sh. I. Mutairov, K. I. Bushmeleva

Surgut State University, Surgut, Russian Federation

Abstract: We conducted a comparative analysis of keyword extraction methods based on statistical and neural network approaches, including YAKE and transformer-based language models such as BERT. We evaluated their accuracy, performance, and suitability for different text types. Our primary goal was to give recommendations for optimizing these methods to enhance text processing efficiency as data volumes increase. We performed experiments with real scientific texts.

Keywords: natural language processing, keyword extraction, YAKE, BERT, accuracy, performance.