Mandy Barrett Korpusik, MIT. “Deep Learning Models for Spoken Dialogue Systems”


Position:  PhD Candidate

Current Institution:  Massachusetts Institute of Technology

Abstract:  Deep Learning Models for Spoken Dialogue Systems

Personal digital assistants such as Siri, Cortana, and Alexa must translate a user’s natural language query into a semantic representation that the backend can then use to retrieve information from relevant data sources. For example, answering a user’s question about the weather requires querying a database with information about the weather at a given time in a specific location. In my work, we are investigating deep learning techniques for performing such a semantic mapping from human natural language to a structured relational database. Specifically, we have explored convolutional neural network architectures for learning a shared latent space where vector representations of natural language queries lie close to embeddings of database entries that have semantically similar meanings. To assess our technology, we have applied these techniques to the nutrition domain and built a full system prototype for diet tracking on the iOS platform.

Mandy Korpusik a master’s degree from MIT in 2015.  She is currently pursuing her PhD in the Spoken Language Systems group at MIT, advised by Dr. Jim Glass as her advisor. she received a bachelor’s degree in electrical and computer engineering from the Franklin W. Olin College of Engineering in 2013. She received the NDSEG fellowship in 2015 and the best-rated poster presentation award for demos at the Spoken Language Technology (SLT) workshop in 2014. Her primary research interests include natural language processing and spoken language understanding for dialogue systems, and she previously worked on user intent detection, semantic tagging, and predicting user purchase behavior.