Yuyi Qu

I am a fourth year undergraduate studying Computer Science at the University of Michigan. I'm currently doing research in CRO+MA Lab.

See my work


Crowdsourcing the Generation of More User-Efficient Clarifications in Conversational Systems
  • Implemented a hybrid workflow that optimizes over information gain and answerability.
  • Compared different levels of presentation of the computer’s knowledge and found that entropy with ranking is most informative and interpretable for humans.
  • Collected synthetic end user data; Designed and conducted an experiment on Amazon Mturk.
Conversational AI - Travel Plan Assistant
  • Built a conversational AI that simulates a travel agency to generate travel itineraries for users.
  • Trained an intent classifier and slot extraction classifiers using end-to-end recurrent and bi-direction LSTM neural network with crowdsourced and manually labeled data.
  • Integrated Clinc conversation AI platform into interaction among frontend, backend server, business log server, Google ASR, Google firebase and other third party APIs like TripAdvisor.
Active Learning with Rule Creation
  • Introduced a hybrid approach, using computers and humans, that generates concept-based rules by leveraging people’s semantic knowledge to reduce labeling efforts for large text dataset.
  • Clustered the data into keyword-based patterns using Latent Dirichlet allocation (LDA).
  • Classified examples by computing cosine similarity with glove word embedding and setting a threshold.
User-Generated-Content Badge Design with Social Status Valuation
  • Calculated optimal mechanism to maximize user contribution to UGC websites.
  • Interpreted realistic significance of concave, convex, linear and arbitrarily-shaped status function.
Website Classifier
  • Processed raw HTML files collected by Censys and manually labeled categories for 200 websites.
  • Implemented a classifier to analyze the composition of the Internet and how it changes over time.


Contact Me