Unsupervised extraction of local and global keywords from a single text

Anonymous

Abstract: We propose an unsupervised method to extract keywords from a single text. It is based on spatial distribution of words and the response of this distribution to a random permutation of words. The method allows inference of two types of keywords: local and global. Several classic literature texts demonstrate that such a classification of keywords is meaningful, and that this method significantly outperforms existing methods (such as YAKE and LUHN) in terms of keyword extraction. Additionally, it is language-independent, applies to short texts (e.g. scientific papers) and uncovers basic themes in texts. Yet another keyword extraction scheme is proposed, but it applies only to texts with many chapters. It is less efficient than the previous one, and is formally similar to metrics used to evaluate scientists (h-index).

Paper Type: short

Research Area: Efficient Methods for NLP