The use of fuzzy ontologies in the clustering of bibliographic information

Abstract

This article presents an approach to clustering short texts using a fuzzy ontology. We propose a modification of the TF-IDF model for vectorization of short texts using a fuzzy ontology. A fuzzy ontology determines the degree of membership between the terms of the subject area. The paper presents a comparison of the efficiency of 4 types of clustering (K-means, MiniBatchKMeans, DBSCAN, Agglomerative) and 3 types of short text vectorization (Bag of Words, Word2Vec and modified TF-IDF). The most effective was the use of K-means and modified TF-IDF for short texts from the news portal. The second set of experiments consisted in clustering texts of abstracts of scientific articles from the elibrary portal. The results of the experiments will be used to create new scientific groups and expand existing scientific groups on topics.