From machine learning to classroom learning - mobile vowels and the Russian preposition v ‘in(to)’

Aug 4, 2026

Upload by Violet

The document reports a machine learning study on mobile vowels in the Russian preposition v ‘in(to)’. A neural network predicts whether the surface form uses v or vo with 97.4% accuracy on a dataset of 18,207 preposition-word samples. A decision tree is then used to derive three learner-oriented rules, while further analysis refines their pedagogical usability. The work supports earlier claims about mobile vowels preventing identical-segment sequences and introduces the Word Onset Hierarchy to assess phonological feature importance, emphasizing gradient and asymmetric behavior. Machine learning is positioned as a helpful tool for language pedagogy for difficult Russian grammar areas.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages17

Reading Time43 min

AccessFree to download

Abstract

Introduction

A machine learning experiment

From machine learning to classroom learning: mobile vowels and the Russian preposition v ‘in(to)’ Tore Nesset and Kevin Xavier UiT The Arctic University of Norway [Tore.nesset@uit.no](Tore.nesset@uit.no), [kxa000@post.uit.no](kxa000@post.uit.no) Abstract The present study reports on a machine learning experiment concerning mobile vowels in the Russian preposition v ‘in(to)’. It is shown that a neural network is able to predict mobile vowels in 97.4% of the cases in our dataset, and a decision tree is used to extract a set of three rules that a language learner can use to achieve nearly the same level of accuracy. We argue that these rules are valuable from the perspective of language pedagogy, but that some adjustments are necessary in order to make the rules simpler and more precise. Our study lends support to earlier analyses which emphasize the capacity of mobile vowels to prevent sequences of identical segments. We advance the Word Onset Hierarchy, which enables us to evaluate the relative importance of phonological features for mobile vowels and highlights the gradient and asymmetric nature of mobile vowels. It is suggested that machine learning represents a valuable tool for language pedagogy, not only for mobile vowels, but also for other areas of Russian grammar that are challenging for students of Russian as a foreign language. 1. Introduction Students of Russian as a foreign language discover very early that some words have forms with vowels that disappear in certain environments. Thus, in (1) the pronoun meaning ‘whole, all’ has a vowel in the stem (ves’), while in (2) this vowel is not present (vsem) . The preposition meaning ‘in(to)’, on the other hand, has a vowel in (2), where it is realized as vo. This vowel is not found in (1) , where the preposition is realized as v:1 1 These examples are from the Russian National Corpus ([www.ruscorpora.ru](www.ruscorpora.ru)) . (1) Ja vljublen v ves’ mir. (Kataev 1980–1981)‘I am in love with the whole world.’ (2) Roditeli vo vsem mire žalujutsja na to, čto detej nevozmožno otorvat’ ot komp’jutera. (“Daša” 2004) ‘Parents all over the world complain that it is impossible to drag kids away from their computers.’ Such vowel~zero alternations known as “mobile vowels” represent a major challenge for learners of Russian as a foreign language, and many instructors have received the question “what is the rule?” – a far from trivial question to answer.2 Mobile vowels are not only challenging for language pedagogy, they also represent one of the classic theoretical problems of Russian phonology. Simplifying somewhat, the theoretical literature addresses the following questions: Are the vowel~zero alternations we can observe the result of epenthesis or deletion (see Bethin 1998: 210– 211, Scheer 2011 and 2019: 202–203 for discussion)? Are there one or two underlying vowels that produce vowel-zero alternations (e.g., Iosad 2020)? What is the conditioning environment for the alternations (e.g., Klapper 1993, Gouskova 2012, Gouskova and Becker 2013 , Becker and Gouskova 2016, Linzen, Kasayanenko and Gouskova 2013)? Prepositions figure prominently in the theoretical literature about mobile vowels, since they are relevant for the relationship between prepositions and prefixes – a much debated issue in Russian morphosyntax (e.g., Matushansky 2002, Steriopolo 2007, Gribanova 2009, Blumenfeld 2012) . In this article, we consider the conditioning environment of mobile vowels from the point of view of language pedagogy. We concentrate on the preposition v ‘in(to)’ and ask whether a machine learning experiment can help us formulate rules that can be useful in the classroom. Our contribution can be summarized as follows. First, we show that a neural network is able to learn to choose between v and vo; the model predicts the correct variant in 97.4% of the cases in our dataset of 18,207 prepositionword samples. Second, we develop a decision tree and argue that the model

FAQs

What linguistic problem does the study focus on?

The study focuses on mobile vowels in the Russian preposition v ‘in(to)’, where the vowel alternates between v and vo depending on the surrounding word environment.

How accurate is the neural network in predicting v vs. vo?

The neural network predicts the correct variant in 97.4% of cases in the dataset of 18,207 preposition-word samples.

How are machine-learning results used for teaching?

A decision tree extracts three simple rules that learners can apply, and the authors argue that these rules need adjustments to become simpler and more precise for classroom use.

Aug 4, 2026

Upload by Violet