Features of using neural networks in the development of productive lexico-grammatical skills in the Catalan language for students of Spanish

The aim of this study is to determine the potential and limitations of using neural networks in the development of productive lexico-grammatical skills in the Catalan language for students of Spanish. The article analyzes the features of integrating artificial intelligence programs (using Sesame AI, ChatGPT и GPTs the Language Teacher Mr. Smith as examples) into the process of students' independent speech practice. The main focus is on examining the functions of neural networks, such as generating exercises, modeling dialogues, and providing instant feedback. The advantages of these tools for expanding the practice of learning a less commonly taught language with a limited amount of classroom instruction are shown. The scientific novelty of the research lies in a comprehensive assessment of the effectiveness of using neural networks in the specific context of teaching the Catalan language. As a result, promising directions (role-playing games, creation of contextual tasks) are identified, as well as existing limitations, including the anglocentricity of models and possible errors in the generated materials. The research substantiates the need for adapting digital content to improve its quality and compliance with the goals of the educational program. Neural networks are considered as a means of supporting traditional teaching methods, contributing to the development of motivation, autonomy, and communicative competence of students.