Florian Jaeger

Affiliation: 
University of Rochester
Talk Title: 
Why and how is speech processing so flexible?
Event Type: 
Colloquium
Semester: 
Spring 2017
Date: 
Friday, April 14, 2017, 3:30 pm - 5:00 pm
Location: 
SAC 305
It is now quite broadly recognized that language understanding draws on implicit expectations based on the statistics of previously experienced input. These expectations facilitate efficient goal-oriented processing, and more robust understanding in noisy environments. At the same time, many of the relevant input statistics aren't stationary--they change depending on who we talk to, the current topics, etc. Human communication is a highly interactive process. How then do listeners and talkers navigate the resulting variability in the input statistics over time?
 
I'll first provide a brief overview of how to think about this problem in terms of ideal adapters, extension of probabilistic inference models that describe ideal behavior under varying statistics in the world. I then present examples from speech perception and speech production that test predictions of these frameworks. The goal is to illustrate how these models, and the types of experiments the encourage, can help us understand why and how the systems and representations underlying language processing manage to flexibly adapt to different situations.
 
Selected refs:
Buz, E., Tanenhaus, M. K., & Jaeger, T. F. (2016). Dynamically adapted context-specific hyper-articulation: Feedback from interlocutors affects speakers’ subsequent pronunciations. Journal of Memory and Language. http://doi.org/10.1016/j.jml.2015.12.009
 
Fine, A. B., Jaeger, T. F., Farmer, T. a., & Qian, T. (2013). Rapid expectation adaptation during syntactic comprehension. PloS ONE, 8(10), e77661. http://doi.org/10.1371/journal.pone.0077661
 
Kleinschmidt, D. F., & Jaeger, T. F. (2015). Robust speech perception : Recognize the familiar , generalize to the similar , and adapt to the novel. Psychological Review, 122(2), 148–203. http://doi.org/10.1037/a0038695