“Guess Who’s Coming to Dinner: a Bayesian Approach to Modeling Perspective”
Abstract: Perspectival aspects of meaning are some of the most challenging dimensions of natural language for artificial learners because perspective, or point-of-view, is situation-dependent, dynamic, and grounded in the physical world.
I seek to model the reasoning process that conversation participants use to select and interpret context-sensitive expressions that are interpreted relative to a point-of-view. For instance, the perspectival motion verb ‘come’ requires a perspective-holder to be located at the destination of motion. When a listener hears a sentence like ‘Thelma is coming to the zoo’, how do they decide whose perspective the speaker is using?
I model the listener’s interpretative process as a Bayesian inference process: listeners reason jointly about the speaker’s intended meaning and their adopted perspective using a mental model of how the speaker selected the utterance. I generate predictions from simulations run in the WebPPL probabilistic programming language and provide empirical evidence from crowdsourced behavioral experiments in support of a key prediction of the model: that listeners simultaneously consider multiple perspectives.
Bio: Carolyn Anderson is a computational linguist who studies how point-of-view is encoded in natural language. Her research applies Bayesian inference techniques to model how conversation participants reason about each other’s spatial and mental perspectives. She also works on techniques for automatic speech recognition in low-resource contexts. Before starting her PhD in Linguistics at the University of Massachusetts, Amherst, Carolyn worked on technology for language revitalization as a Fulbright scholar at McGill University.
All are welcome. A reception for attendees will be held at 3:30 p.m. in CS 150