Using computational modeling and behavioral experiments—with an empirical focus on the under-resourced languages of Pakistan—Dr. Paramore examines what linguistic and social information is encoded in speech representations, how those representations support speech production and perception, and how insights from human cognition can inform the development of more effective automatic speech recognition systems.

Across these projects, he asks a common set of questions: What information is encoded in speech representations? How abstract or detailed can those representations be? How are they learned? And how can insights from human language inform the design of better AI models?

A central commitment of his research is supporting understudied languages through documentation, community collaboration, and the development of language technology. To this end, he works closely with speakers and communities in Pakistan to document and analyze interesting linguistic features, to develop language resources, and to create computational tools that contribute to language preservation and revitalization.