A Quantitative Analysis of Unsupervised Keyword Extraction Methods

This code is part of a Bachelor thesis at Humboldt University Berlin.

This code will compare commonly used unsupervised keyphrase extraction methods, which are structured in three areas: Statistical methods (TF-IDF, YAKE, RAKE), mathods that are based on graphs (SingleRank, TextRank), and deep learning methods (KeyBERT). The latter uses language-specific pre-trained models, in this case for German, whereas the statistical and graph-based methods need no training and are language unspecific.