Time: June 4, 4:25-5:25pm Purpose: In the spirit of a workshop, we are planning an informal panel discussion. The goal is to get together and figure out what important issues in SSL-NLP need more attention, what lessons should be shared, etc. The material here is still tentative and subject to change. Please let us know (email ssl.nlp2009@gmail.com) if you have any suggestions! Panelists: - Hal Daume (University of Utah) - Andrew Goldberg (University of Wisconsin - Madison) - David McClotsky (Brown University) Invited Position Paper: - Sajib Dasgupta and Vincent Ng (UT Dallas). Discriminative Models for Semi-supervised Natural Language Learning - Hal Daume (U of Utah). Semi-supervised or Semi-unsupervised? Topics Discussed: - Should there be a shared task for SSL-NLP, and if so, how should the data and evaluation be set up? - What do you think are the open problems in semi-supervised learning for NLP? What should researchers focus on? - Are there any emerging applications where SSL-NLP will be an important enabler? - Human language learning and semi-supervised learning -- what are the connections that are worth investigating? - Online learning and lifelong learning. |
|
|
Panel Discussion |