In this talk, I will show how we can combine statistical NLP methods and sociolinguistic theories to the benefit of both fields. I present research on large-scale statistical analysis of demographic language variation to detect factors that influence performance (and fairness) of NLP systems, and how we can incorporate demographic information into statistical models to address both problems.
Sociolinguistics has long investigated the interplay of demographic factors and language use, and the same factors are also present in the data we use to train Natural Language Processing (NLP) systems.
NLP models, however, are based on a small demographic sample and approach all language as uniform. As a result, NLP models perform worse for demographic groups that differ from the training data. This bias harms performance and can disadvantage entire user groups. I will show how adding demographic information to NLP models can improve performance and create fairer systems for everyone.