Organizer

Hang Li (Microsoft Research Asia, Beijing)

Recently, a new trend has arisen in the field of Natural Language Processing (NLP): the development of machine learning technologies that use both labeled and unlabeled data for training. Methods that have been proposed under this paradigm include co-training, EM learning, transductive learning, and other semi-supervised learning techniques. For many NLP tasks, existing data are by their nature unlabeled and manually labeling them is prohibitively expensive. Effective utilization of both unlabeled and labeled data in learning is also a challenging but important issue. The goal of this thematic session is to bring together researchers working on this issue from different perspectives, in order to share their latest research results and to discuss future directions. We think that this session will advance research not only in exploiting unlabeled data but also in other natural language learning issues.

Submission

The submission procedures and deadline for thematic sessions are the same as those for the main conference

Program Committee

Steven Abney, University of Michigan, USA
Ido Dagan, Bar Ilan University , Israel
Hang Li, Microsoft Research Asia, China
Chin-Yew Lin, University of Southern California, USA
Yuji Matsumoto, Nara Institute of Science and Technologies, Japan
Hwee Tou Ng, National University of Singapore, Singapore
Cheng Niu, Cymfony, USA
Miles Osborne, University of Edinburgh, UK
David Pierce, University of Buffalo, USA
Anoop Sarkar, Simon Fraser University, Canada
Eiichiro Sumita, ATR, Japan
Maosong Sun, Tsinghua University, China