Speaker adaptation for robust communication

University Of Rochester, Rochester NY

Investigators

Linked publications & trials

Abstract

? DESCRIPTION (provided by applicant): Language provides speakers with the ability to encode the same message in different ways, so a critical question is what drives the decisions speakers make during linguistic encoding? The current view of the language production system argues that speakers prefer to encode their message based on what is easiest, based on evidence that speakers are sensitive to and prefer more easily produced forms over alternatives. However, what is easy for a speaker to produce is not always easy for a listener to comprehend. An open question is to what extent speaker decisions might also reflect a bias for forms that facilitate communication. Recent findings suggest that speakers modulate their productions based on changes to contextual confusability, something not predicted by the view that speakers ignore the intelligibility of their productions. Tested here is an alternative accoun that argues speakers balance production ease with the goal of robust communication and do so by learning from perceived successes and failures at conveying their intended message to listeners. This proposal seeks to test if the aforementioned contextually driven speech changes do facilitate communication using a novel simulated-communication paradigm. This proposal further seeks to test a novel prediction of the robust communication account: that speaker decisions change based on the perceived successes and failures of previous attempts at communication. Preliminary evidence supports this prediction and the project proposed here will further test and refine this account of speaker behavior. The completion of this proposal will lead to insights on if and how the speech production system is organized to achieve communicative goals by adapting to past successes and failures.