Artificial Intelligence-Based Corpus Analysis of Linguistic Identity in Contemporary Pakistani Fiction

Abstract

This study examines how linguistic identity is constructed in contemporary Pakistani fiction written in English, using an artificial intelligence (AI)-assisted corpus-analytic framework. A purpose-built corpus of 24 novels and short-story collections published between 2000 and 2023 (approximately 2.94 million running words) was compiled, digitized, and annotated. The corpus was analyzed through an integrated pipeline that combined classical corpus-linguistic techniques (keyword extraction, collocation, concordance, n-gram profiling) with contemporary natural language processing methods (transformer-based part-of-speech tagging and named entity recognition, language identification for code-switching detection, sentence-embedding clustering, BERTopic modelling, and aspect-level sentiment and stance classification). Findings indicate that linguistic identity in the corpus is realized through five interlocking systems: (i) dense lexical borrowing from Urdu, Punjabi, Sindhi, and Pashto, concentrated in the semantic fields of kinship, food, religion, clothing, and address; (ii) intra-sentential code-switching that functions as an index of intimacy, authenticity, and social positioning rather than as a communicative deficit; (iii) a distinctive collocational grammar around identity nouns such as 'home', 'country', 'people', and 'name'; (iv) pronominal patterning in which first-person plural forms carry the burden of collective and national self-definition; and (v) evaluative and stance patterns that are markedly ambivalent, combining affiliation with critique. Topic modelling recovered nine stable semantic themes, of which migration and belonging, religion and moral order, and urban violence were the most productive of identity language. The study argues that AI-based corpus analysis makes visible the systematic, distributional character of literary identity work that close reading can only sample, and it proposes an integrated AI-corpus model of linguistic identity for Anglophone postcolonial fiction. Methodological limitations, validation procedures, and directions for future research are discussed.