A Reconsidered Language Model Approach to the Problem of Verb Aspectual Pairs in the Russian Language

Background. The problem of verb aspectual pairs in Russian studies remains relevant and still causes discussion, despite the large number of studies conducted. Triplets attract intense interest due to their special position in the system of aspectual forms and the implicitness of an aspectual pair of verbs in the triple. Another fundamental issue that causes a lot of discussion is the so-called hypothesis of empty prefix. Purpose. The purpose is to obtain objective quantitative evaluation of the semantic distance between the verbs in triplets: the basic imperfective verb, the perfective verb, and the secondary imperfective verb. The main problem we solve is which of the two imperfective verbs is semantically closer to the perfective verb, i.e. which pair is aspectual. Materials and Methods. So far, there have been no operational methods for determining verb aspectual pairs. With the advent of large text corpora and Language Models technology, it has become possible to evaluate the semantic similarity of words using rigorous methods. To solve the issues, the computational technique of embeddings (vector semantics) is used. The present study covers 440 triplets. The data on their frequency and collocations were taken from the Russian subcorpus of the Google Books Ngram database. Results. For both language models applied, statistical test shows greater semantic similarity for the pairs of perfective and secondary imperfective than for the pairs of basic imperfective and perfective. According to one or both of the models, for most of prefixes considered, the average distance between the pairs of basic imperfective and perfective is statistically significantly greater than between the pairs of perfective and secondary imperfective. Conclusion. The article presents quantitative analysis of identity of semantics in aspectual pairs of triplets. Secondary imperfective verb form is shown to be statistically closer to the perfective verb than basic imperfective form. This is an argument in favor of the point of view that secondary imperfective verbs form aspectual pairs with perfective verbs. Also, the models applied show that prefixes contribute to the semantics of the perfective verbs. This fact stands against the hypothesis of empty prefixes.