Modeling Russian Verbs of Motion: An Analogical Account

This paper presents research into verb of motion (VOM) constructions in Russian. These constructions are difficult since they involve (i) selection of an appropriate verb (with possible prefixation); (ii) selection of an appropriate preposition; and (iii) selection of an appropriate case for marking the prepositional object. A brief sketch of relevant literature frames the problem. We then discuss how a few thousand instances of VOM usage were extracted from an online tagged corpus of Russian literature. The usage instances were then vectorized using a combination of lexical and semantic class features via automatic, semiautomatic, and hand-coded methods. The instance base was then processed via the analogical modeling paradigm to account for predictability of the case, the preposition, and both simultaneously. Discussion of the results and possible future research directions then follows.