Most research focuses on generalizing or assuming a one-size-fits-all language framework. Rios, however, will apply the grant funding to develop NLP models that are specific to geographic populations, ethnic communities and other population-specific groups.
“NLP is about developing tools to understand language, but we do not have language without people,” explained Rios, who is also the associate director of the UTSA Cyber Center for Security and Analytics. “My specific focus is developing NLP tools for biomedical and social applications, which I feel can have a real-world impact and hopefully improve people’s lives.”
Rios’s project has three objectives: determine methods to detect correlations between community language and NLP model performance; develop a tool to help community leaders find potential harms of specific NLP models and help them decide if those harms are relevant to that community; and to find processes and opportunities to improve NLP model performance.
“I’m interested in understanding how these models can impact people, and what biases exist within them,” Rios said. “There is a lot of hype surrounding NLP and artificial intelligence, but I think it’s important to avoid hype and to understand the potential harms caused by NLP models when they are hype.”
Rios comes to this project with a background and multidisciplinary research interests in NLP, machine learning and computational social science. He received his introduction to NLP during his computer science doctoral program at the University of Kentucky, where his faculty advisor was Ramakanth Kavuluru, an experienced NLP researcher. NLP’s numerous applications and its impact on people drew Rios to the field.