Using Machine Learning to Study Neural Representations of Language Meaning


#1

Title: Using Machine Learning to Study Neural Representations of Language Meaning
Date: Thursday, June 15, 2017
Time: 12:00 PM Eastern Daylight Time
Duration: 1 hour

SPEAKER
Tom Mitchell
Professor at Carnegie Mellon University

Resources
Register for Live Presentation


#2

Could this type of research where you are scanning the brain to find areas of activity generated by visual or language stimuli potentially be useful for “reading minds” or figuring out what people are actually thinking about when you have access to live scans of their brain activity?


#3

For Tom Mitchell:
Do you have any insights on your work wrt paraphasias? When speaking or hearing the wrong word, it is usually close to the intended or spoken one semantically, phonetically, morphologically, or on some other level (presumably “close” to each other in a vector space on the corresponding CNN layer).
It would be interesting to map specific paraphasias caused by localized, identified brain lesions to the corresponding voxels in your models, and see how close to the injured brain your classifiers behave.


#4

Interesting material. Perhaps this approach would be useful to learn about diseases or syndromes as well.