Automated Compilation of a Corpus-Based Dictionary and Computing Concreteness Ratings of Russian

Abstract

© 2020, Springer Nature Switzerland AG. The article presents new method implemented by the authors to generate dictionaries of concrete/abstract words for Russian. The method based on pretrained word embeddings computes concreteness ranking defined as a function of similarity between word vectors and the distance between a word in question and the ‘seed’ of concrete/abstract words. Implementation of the method resulted in generating the Russian Dictionary of Concreteness/Abstractness with concreteness ratings. Context information encoded by word embeddings useful to rank words by their concreteness score. The resulting C/A rankings strongly correlate with human experts’ assessments.