Research interests

Our research goal is to enable machines to turn massive data into knowledge in a label-efficient and robust way. Revolving around this goal, we study the following topics in data mining and machine learning:
- Low-resource text mining
- Time series and sequential data analysis
- Learning with less labeled data
- Probabilistic modeling and uncertainty quantification

Representative publications

Please find below some of our representative works:
- Weakly-Supervised Neural Text Classification, CIKM 2018
- TaxoGen: Unsupervised Topic Taxonomy Construction by Adaptive Term Embedding and Clustering, KDD 2018
- SenseGAN: Enabling Deep Learning for Internet of Things with a Semi-Supervised Framework, IMWUT 2018
- RDeepSense: Reliable Deep Mobile Computing Models with Uncertainty Estimations, UbiComp 2018
- Doc2Cube: Allocating Documents to Text Cube without Labeled Data, ICDM 2018
- Spatiotemporal Activity Modeling Under Data Scarcity: A Graph-Regularized Cross-Modal Embedding Approach, AAAI 2018
- Bridging Collaborative Filtering and Semi-Supervised Learning: A Neural Approach for POI Recommendation, KDD 2017
- Regions, Periods, Activities: Uncovering Urban Dynamics via Cross-Modal Representation Learning, WWW 2017
- GMove: Group-Level Mobility Modeling Using Geo-Tagged Social Media, KDD 2016
- GeoBurst: Real-Time Local Event Detection in Geo-Tagged Tweet Streams, SIGIR 2016

Ongoing projects

Low-Resource Text Mining

Turning text into knowledge with weak supervision.

Multimodal Data Analytics

Multimodal learning on text-rich spatiotemporal data.