DEVELOPMENT OF A NEURAL NETWORK FOR CLASSIFICATION OF GENERATED TEXTS IN EDUCATIONAL INSTITUTIONS

Alekseeva Ekaterina, Trushin Stepan

This article discusses the features of recognition of generated texts by language models. The architectures for vector representations, as well as the architectures of recurrent networks that are most suitable for working and analyzing generated texts, are given. An open dataset from Kaggle was used to train the neural network. An algorithm based on ELMo and fully connected networks for text classification has been developed. The input data has been preprocessed. An example of the algorithm's operability is presented using the example of school essays. It is concluded that the developed neural network effectively classifies texts and provides significant support in maintaining academic integrity and will serve as a reliable tool for educational institutions.

Keywords:neural networks, artificial intelligence, deep learning, ELMo, RNN, generated text, development, algorithm, language models.