Research work · 2023

Sociolinguistic Perspectives on Machine Learning with Text Data

AJ Alvero

The Oxford Handbook of the Sociology of Machine Learning

Abstract

Abstract Digitized text has become a popular form of data for sociologists. But text is also the product of many different social and linguistic processes, a topic traditionally examined by sociolinguists in the case of spoken language. This chapter presents a sociolinguistic perspective as a methodological framework when using machine learning to analyze text data. The chapter describes sociolinguistic theories and approaches to studying language while also highlighting ways that this literature tends to differ from traditional sociology. To exemplify this approach, the chapter analyzes a corpus of college admissions essays written by Latinx-identifying applicants using two popular machine-learning-based methods: topic modeling and word embeddings. Similar to variation in spoken language among constituent groups and ethnicities within the panethnic category (e.g., Mexican, Cuban), written language also varies by subgroup in meaningful ways. A sociolinguistic perspective for computational text analysis could spur theoretical and empirical insights.

DOI: 10.1093/oxfordhb/9780197653609.013.15