My ASLP journal article on asking followup clarification questions just got accepted, and I'll be presenting at ICASSP in the UK next week! In addition, we'll be presenting two abstracts at the Nutrition 2019 conference in June! We're currently recruiting more participants for an expanded pilot study with our Tufts nutritionist collaborators on the spoken food logging system. Contact me if you're interested, or download Coco Nutritionist from the Apple Store to try it out!
I completed my Master of Science degree in June 2015 and will be graduating with my PhD from MIT this June! My primary research interest is natural language processing (NLP), especially spoken language understanding and dialogue systems. For my PhD, I designed novel convolutional neural network models that perform semantic tagging and mapping of natural language meal descriptions to semantically similar USDA food database entries. I also demonstrated that asking followup clarification questions with deep reinforcement learning boosts food recall performance by narrowing down the search space. I showed that these neural convolutional models generalize to other tasks beyond nutrition, including semantic tagging of user queries in flight and movie booking domains, as well as dialogue state tracking for restaurant booking conversations. I am interested in developing conversational agents that use recurrent neural network models to converse naturally with humans, and reinforcement learning to adapt online to users in order to personalize each interaction.
Deep Learning for Database Mapping and Asking Clarification Questions in Dialogue Systems Mandy Korpusik, James Glass
IEEE Transactions on Audio, Speech and Language Processing, 2019.
Deep Learning for Spoken Dialogue Systems: Application to Nutrition Mandy Korpusik
Ph.D. Thesis, MIT Department of Electrical Engineering and Computer Science, June 2019.
A Food Logging System for iOS with Natural Spoken Language Meal Descriptions Mandy Korpusik, Salima F. Taylor, Sai Das, Cheryl Gilhooly, Susan B. Roberts, James Glass
Testing the Validity of a Natural Spoken Language Application for the Self-monitoring of Daily Dietary Intake Rachel E. Silver, Mandy Korpusik, Salima F. Taylor, Sai Krupa Das, Cheryl H. Gilhooly, James Glass, Susan B. Roberts
April 2016, Service Award. Sidney Pacific Graduate Residence.
April 2015, Fellowship. National Defense Science and Engineering Graduate (NDSEG).
March 2015, Honorable Mention. National Science Foundation (NSF) Graduate Research Fellowship.
December 2014, Best Rated Poster Presentation Award: Demos. Spoken Language Technology Workshop (SLT), South Lake Tahoe.
Korpusik, M. et al. A System and Method for Semantic Mapping of Natural Language Input to Database Entries via Convolutional Neural Networks. U.S. Patent Application 15/922394, filed March 15, 2018. Patent Pending.