The Semantics of Phrases and Sentences in the Human Brain
Semantic composition is the act of creating higher order meaning by combining single word meanings, which is a crucial part of communication and comprehension. Though we have some understanding of how people process and represent the semantics of single words, we do not yet fully understand the neural processes that govern semantic composition. In this talk, I will describe our recent progress towards decoding the semantic content of phrases and sentences from Magnetoencephalography data (high temporal resolution brain activity recordings). Inspired by previous work, we represent each word's meaning with a vector of statistics culled from a large text corpus. We explore how different corpus statistics, sentence structure, and the semantic predictability of the sentence impact our ability to decode these vectors from the brain activity.
Alona Fyshe is a PhD Candidate at Carnegie Mellon University in the Machine Learning Department. Alona received her BSc and MSc in Computing Science from the University of Alberta where she developed new techniques to improve protein localization prediction. After her MSc, Alona worked as a software engineer at Google's Pittsburgh office applying machine learning techniques to improve online shopping. For her PhD thesis, Alona is using machine learning to leverage large amounts of text and neuroimaging data in order to understand how the brain combines words to create higher-order meaning.
Find more undergrad events on our internal portal at https://my.cs.ubc.ca.
This event's address: https://my.cs.ubc.ca/event/2014/09/talk-alona-fyshe-carnegie-mellon-university