My students Yohan, Dongwoo, and I visited Princeton two days ago. (Now we’re at Carnegie Mellon, which may be the topic of my next blog post.) Our original plan was just to visit CMU for some meetings about a collaboration effort. But we thought it would be a much better plan to add another fruitful meeting since we are spending thousands of dollars on plane tickets and many many hours traveling from Korea to the US east coast.

Since we have been working with topic models for the last six months, I thought David Blei at Princeton would be the best person to visit. He accepted my request, and on top of that, invited us to give a talk at his machine learning reading group. Although we don’t yet have fancy results, we gave a presentation that summarizes our research goals, preliminary results, discussion questions, and future plans. We also had a separate meeting with Dave which led to helpful discussions and neat ideas. Dave was a gracious host and answered lots of our questions even though he had a very jam-packed day because it was the visit day for Princeton CS new admits. It was a great experience for us, and we will follow up on our discussions with Dave by trying out some new ideas for LDA evaluation and topic-sentiment model.

Here are the slides from the presentation. (Yohan and I made & delivered the talk, Dongwoo and Hyunjong contributed to the results in the talk.)
Talk @ Princeton, March, 2010

The readings associated with the presentation are:

J. Chang, J. Boyd-Graber, S. Gerrish, C. Wang, and D. Blei.    Reading tea leaves: How humans interpret topic models. Neural Information Processing Systems, 2009.

C. Lin and Y. He. Joint Sentiment Topic Model for Sentiment Analysis. CIKM 2009.

If you’re interested in learning about LDA, Dave’s tutorial is great.