2023, Vol. 14, No. 3. - go to content...

Permanent address of this page - https://sfk-mn.ru/en/54flsk323.html

Метаданные этой статьи так же доступны на русском языке

Full article in PDF format (file size: 523.2 KB)

For citation:

Gorodishchev A.V., Uskova S.V., Khodenkova E.V. [Processing of the Russian language by artificial intelligence models in the formation of a digital avatar] World of Science. Series: Sociology, Philology, Cultural Studies, 2023, Vol. 14, No. 3. Available at: https://sfk-mn.ru/PDF/54FLSK323.pdf (in Russian).

Processing of the Russian language by artificial intelligence models in the formation of a digital avatar

Gorodishchev Alexey Viktorovich
National University of Science and Technology «MISIS», Moscow, Russia
E-mail: agorodischev@gmail.com
ORCID: https://orcid.org/0000-0003-2224-3794
RSCI: https://elibrary.ru/author_profile.asp?id=1050008
SCOPUS: https://www.scopus.com/authid/detail.url?authorId=57221043983

Uskova Svetlana Viktorovna
Reshetnev Siberian State University of Science and Technology, Krasnoyarsk, Russia
E-mail: s_us78@mail.ru
ORCID: https://orcid.org/0000-0003-0074-8417
RSCI: https://elibrary.ru/author_profile.asp?id=673911

Khodenkova Elga Vladimirovna
Reshetnev Siberian State University of Science and Technology, Krasnoyarsk, Russia
E-mail: ellie4ka@mail.ru
ORCID: https://orcid.org/0000-0002-8747-2742
RSCI: https://elibrary.ru/author_profile.asp?id=842235
SCOPUS: https://www.scopus.com/authid/detail.url?authorId=57209853221

Abstract. In this paper, the authors proceed from the fact that in the modern world online discourse is not only rapidly changing, but also dictates the formats of communication. If in the first decade of the XXI century in the online space genre, linguistic and extralinguistic features of the functioning of discourse as a whole developed according to the rules of communication of real people, then in the third decade the question arose of the applicability of the laws of language development to the technologies of generating texts and speech of digital characters.

The aim of the study is to compare the ways of processing words and expressions of the Russian spoken language and slang with various models of artificial intelligence (AI), revealing its ability to recognize the ambiguity of a word, stylistic connotations, cultural meanings and moral and ethical limitations. Firstly, the relevance of the study is that for the first time, the features of processing specific exclusively Russian language phrases by AI models were compared using the example of three neural network models that are in the public domain, saiga_7b_ggml, RUGPT3.5, ChatGPT 3.0. Secondly, the issues of AI speech adaptation in the formation of a digital avatar with the preservation of not only the linguistic characteristics of the personality, but also its cultural significance are considered. As a result of the experiment, it was found that the tested neural network models do not recognize national Russian meanings, emotional and stylistic connotations, often creating a feeling of a random choice of the answer. It follows from this that when forming the language of digital avatars, speech moderation is necessary not only at the initial stage of training digital models, but also in the process of changing the context of the conversation depending on the emotional colouring, the tone of the utterance, etc.

Keywords: Russian language; artificial intelligence; language model; neural network; digital avatar, stylistic connotations; Russian colloquial speech

Download article in PDF format