Exploring Code-Switching Dynamics in Urdu-English Multilingual ChatGPT Models: Patterns, Challenges, and Implications

Abstract

This research investigates the code-switching dynamics in the Urdu-English multilingual ChatGPT models aimed at discovering the themes, challenges, and implications. Utilizing text data retrieved from online resources, social media platforms and subject-oriented conversations, code switching will be examined through preprocessing and annotation processes. Algorithms are developed to automatically detect and classify code-switching instances, followed by an in-depth analysis of frequency, distribution, and contextual triggers. The study evaluates the role of ChatGPT in code-switched activities by generating text sets and ranking them based on language identification, syntactic coherence, and semantic consistency. Data evidenced that code-switching is often and that ChatGPT can communicate in different languages. The findings will be helpful in the process of refining AI-based natural language processing systems. The work investigates the more detailed perception of language change in digital environments. It provides a basis for designing more welcoming and culturally considerate communication and media tools.