HSE researchers create new corpus of early child speech in Russian

Researchers at the HSE Centre for Language and Brain have presented RusLan-M, an open multimedia corpus that makes it possible to trace the development of early child speech in Russian from first words to the emergence of complex grammatical constructions. The database contains around 41 hours of video recordings and more than 35,000 child utterances. The new resource will help researchers study more precisely how children acquire Russian and, in the longer term, develop more reliable tools for assessing speech development. The study has been published in Language Resources and Evaluation.

Despite the widespread use of Russian, data on its early acquisition remain limited. The international CHILDES database brings together hundreds of collections covering more than 40 languages, but Slavic languages are significantly underrepresented. At the time the study was prepared, only two corpora of children’s Russian speech were available in CHILDES. The new RusLan-M resource is intended to address this gap by providing researchers with detailed, standardised, and accessible data on how children acquire Russian.

The corpus is based on observations of two Russian-speaking children, Tosya and Yasha. Researchers at the HSE Centre for Language and Brain recorded the development of their speech over an extended period, rather than comparing children of different ages at a single point in time. The collection comprises 288 recordings: Tosya was recorded from the age of 10 months to 3 years and 10 months, while Yasha was recorded from 1 year and 4 months to 3 years of age. The total duration of the video recordings was 2,454 minutes, or around 41 hours, while the number of child utterances reached 35,386.

All recordings were accompanied by transcripts in the international CHAT format and published on TalkBank, an international resource for storing and analysing speech data. When publishing the corpus, the researchers complied with requirements concerning personal data protection and the anonymisation of personal information.

A key feature of the new resource is that it makes it possible to study children’s speech in a natural environment. Unlike laboratory experiments or parental questionnaires, the video recordings capture children’s spontaneous interactions with people close to them and their everyday linguistic environment. This makes it possible to analyse not only what children say, but also the speech context in which their language develops.

‘It was important for us to create a resource that would allow us to see the development of children’s speech not as a set of individual test results, but as a living and continuous process. Thanks to the longitudinal recordings, we can trace how an individual child’s speech gradually changes, which grammatical constructions emerge earlier and which later, and how the linguistic environment influences development. Such open and comprehensively annotated data is particularly important for Russian,’ said Mariia Diachkova, Research Assistant at the HSE Centre for Language and Brain and one of the study’s authors.

Pilot studies have demonstrated how RusLan-M data can be used to analyse the development of morphology and syntax over time. The corpus makes it possible to track the emergence and increasing complexity of specific linguistic constructions, compare different stages of speech development, and formulate new hypotheses about the mechanisms of language acquisition.

According to the authors, the significance of such data extends beyond fundamental linguistics. ‘Corpus-based measures can be used to develop normative benchmarks for speech development at different ages. In the future, this could help specialists distinguish more accurately between individual patterns of language acquisition and potential difficulties, as assessments would be based on data from children’s real spontaneous speech,’ notes Valeriia Lelik, Research Assistant at the Centre for Language and Brain and one of the study’s authors.

The authors emphasise that the new corpus partly addresses the significant shortage of open data on the development of the Russian language.

‘Diary records of child speech have been used as research material in Russian developmental linguistics for decades. Everyone working in this field knows how much effort is required to create corpora of child speech. However, it is important to note that no multimodal corpora of child speech in Russian comparable to RusLan-M—incorporating audio, video, and text and featuring detailed linguistic annotation—had previously been publicly available. Thanks to this large-scale effort, the research community now has access to this valuable tool,’ said Svetlana Dorofeeva, project lead and Senior Research Fellow at the HSE Centre for Language and Brain.

The corpus was created through the joint efforts of a large team of parents, linguists (both researchers and students), and programmers. It brings together first-hand parental experience, academic expertise, and advanced digital technologies.

‘In the longer term, expanding collections of this kind will make it possible to compare the speech development of a larger number of children, identify age-related patterns, and create more accurate models of language acquisition,’ said Olga Dragoy, one of the study’s authors and Director of the HSE Centre for Language and Brain.

RusLan-M also demonstrates how corpus linguistics methods can transform years of observations of children’s speech into a resource for a wide range of research, from theoretical linguistics to psychology and the practical assessment of speech development.