Neural Network Tools for Creating the Literary Texts Based on Transcribed Speech

Khramov, Oleg Stanislavovich
Senior Lecturer, the Humanities and Social Sciences Center, MIPT University, Moscow, Russia
khramov.os@mipt.ru

Potemin, Alexey Pavlovich
Guidance Counselor, the Phystech-Lyceum’ E-learning Center, Moscow, Russia
alex.potemin@kapitsa.center

Abstract
The Russian language—like any other language in the world—consists of the two different linguistic practices: an everyday speech and literary texts. Each of these practices is specific: they are lexically and grammatically different. The using of the direct transcribed speech in literary texts is impossible: the literary editing is required.
This work focuses on the translating speech into literary text in automatic mode; we work with the neural network language model of the GPT 3.5—GPT 4.
Our tool has a range of perspective outcomes. Here and now, the authors are able to create a book based on the recordings of their lectures and talks.
The tool is easy to use: its potential will be demonstrated during the presentation; the audience will have all chances to test this technology.

Keywords: neural network technologies, literary text, live speech