picture

Contact Information

John Blitzer

Postdoctoral Fellow
Computer Science Department
University of California, Berkeley

Email: blitzer@cs.berkeley.edu
My curriculum vitae

What's New

Better Word Alignment with Supevised ITG.  ACL 2009.
Aria Haghighi, John DeNero, Dan Klein, and I develop supervised ITG models that achieve state-of-the-art AER results and a significant improvement in BLEU score over IBM model 4.

Exploiting Bilingual Information to Improve Monolingual Web Search.  ACL 2009.
What do Thomas Hobbes and Han Feizi have in common?  They are both search queries that Wei Gao, Ming Zhou, Kam Fai Wong, and I can improve using bilingual data.

[ about me ]      [ projects ]      [ publications ]      [ presentations ]     

Bilingual Models for Parsing and Entity Recognition

I have received an Intelligence Community Postdoctoral Fellowship for a project on using bilingual data to improve parsing and entity recognition. The ultimate goal is an improved end-to-end translation pipeline. This project is joint work with David Burkett and Dan Klein.


Joint Cross-Lingual Ranking

At Microsoft Research Asia, I worked on applying machine learning to cross-lingual ranking. For many Chinese queries, we can find an accurate English (or other foreign) translation. Our system learns a better ranking function by simultaneously using information from both Chinese and foreign results.


CALO

The CALO project was a multi-institution effort to develop a Cognitive Agent that Learns and Organizes. As part of the UPenn team, I worked on building prototypes for email urgency prediction, summarization and clustering.


Google News

In 2004 I interned with the Google News team. My mentor was Thorsten Brants. I worked on the clustering component, but I think the best way to get to know Google News is to visit the site!


PARC discussion list summarization

During the summer of 2003, I worked at (formerly Xerox) PARC, with the HDI group in the Information Sciences and Technologies Laboratory. My PARC supervisor, Paula Newman, and I built a clustering and summarization system for archived discussion lists, such as email and newsgroups.


Johns Hopkins CLSP workshop

In summer 2002, I participated in a Johns Hopkins Center for Language and Speech Processing workshop on automatic multidocument summarization.