Investigating the Phenomenon of Code-Switching through Computational Linguistics Techniques

Abstract

Code-switching, the use of two or more languages with in a single discussion or utterance is a widespread feature of multilingual communication, mainly in digital environments such as social media and messaging platforms. While sociolinguistic studies have provided a valuable insight into the phenomenon, it remains a complex challenge for computational modeling, especially when dealing with informal structures, grammatical variations, and underrepresented language pairs such as Urdu-English and Hindi-English. This study investigates the phenomenon of code-switching using computational techniques to improved analyze and interpret multilingual language behavior in actual contexts. The research applies advanced Natural Language Processing (NLP) tools including language credentials, part-of-speech tagging, and syntactic parsing on data composed from online interactions. Using annotated corpora and models such as BiLSTM and multilingual BERT (mBERT), the study assesses model correctness and identifies core limitations in processing code-switched text. Furthermore, the study proposes improvements through context-aware strategies that assimilate linguistic insights with machine learning. By connecting computational techniques with language theory, the research contributes to the growth of more inclusive and accurate multilingual NLP tools, with potential applications in sentiment analysis, speech recognition, and machine translation. It also enhances value to the educational and technological landscape of linguistically diverse regions like South Asia.