Click here to Access the Policy Agenda Classifier

The Policy Agenda Classifier is an evolving, online tool created from my public papers on using natural language processing and machine learning to classify the policy activities of the U.S. Congress and the legislative activities of foreign governments.

The tool linked from this web site is based on Thorsten Joachim's SVMlight but its complexity continues to grow. Amazon's Elastic Compute Cloud (EC2) provides the backend processing for the system. The software was developed by myself (machine learning, compute clustering, user interface), Dustin Hillard (machine learning), and Ray Amberg (user interface) and Ray's participation was funded by the Connecting to Congress project, the Congressional Management Foundation, and NSF.

This web site is used by about a dozen project teams working on analysis of political activity. Money collected from the site advertising and donations is used to fund the significant costs of providing this public resource. Contact me if you wish to contribute.

Special thanks to the following people for their contributions:

• Dustin Hillard
• Ray Amberg
• Thorsten Joachims
• Claire Cardie
• John Wilkerson
• Frank Baumgartner
• Bryan Jones
• David Lazer
• Kevin Esterling
• Michael Neblo
• Jamie Callan
• Ed Hovy
• Lillian Lee
• Eric Breck

The papers associated with this project, including portions of the orgiginal algorithms, were advanced with support from the Connecting to Congress Project (NSF grant No. 0429452), and the Congressional Bills Project (NSF Grants No. 00880066, 0111443, 00880061). The views expressed are those of Stephen Purpura and not the National Science Foundation nor the other authors.