Humans Linguistically Align to their Conversational Partners, and Language Models Should Too
- 2024
- ICML 2024
I am a cognitive scientist with a focus in psycholinguistics - the study of speech production and comprehension. I am interested in how speech changes as a function of the context, and specifically, how speakers and listeners adapt their language processing to the characteristics of their particular partner. There is adaptation both on the comprehension side - for example, listeners expect to hear higher-frequency (i.e., easier) words from a 2-year-old than from an adult, as well as on the production side - for example, speakers engage in alignment, when they modulate characteristics of their own speech to converge upon those characteristics of their listener's speech.
I also investigate the use of language production to diagnose cognitive diseases and disorders. Language is an extremely rich signal, and is highly variable, and can be a signal of the speaker's underlying cognitive state. Attentional load, distraction, memory - all affect characteristics of people's speech. Speakers and listeners exploit the massive variability of language in order to improve communication. We hope to exploit that variability to diagnose cognitive disorders.
Please click here to visit my academic website, which is updated more frequently and has more detail about my research and copies of publications.