Jason Rennie's Home Page

Graduate Student (n=0), MIT, AI Lab Assistant Research Scientist, MITRE, 1999 B.S., Carnegie Mellon University, Computer Science and Mathematics, May 1999 Graduated from Great Valley High School, 1995. My (outdated) resume: Postscript, Plain Text.

Metaverse:

jrennie@mitre.org

http://www.andrew.cmu.edu/~jr6b/

Research

• Machine Learning applied to Text.
  I have become quite interested in the problems of classifying text, extracting information from text, retrieving information from text, document clustering and document summarization. I have been involved with the Text Learning Group at CMU. Many of those involved with the Text Learning Group are also active in the WebKB project.

  At one time, I was part of a project with Andrew McCallum, Kristie Seymore and Kamal Nigam at Just Research called Cora, a search engine for Computer Science research papers. While potentially useful to many a researcher, the real interesting part of this project is the automatic self-construction that we were exploring. How does one quickly and efficiently gather documents to index? How does one automatically extract information from documents for display in search results? How does one easily create a hierarchy that represents the underlying semantic structure inherent in the document colleciton? I can't say that we completely solved all of these questions, but we did end up with a very useful search engine in the process.

  I am the author of ifile - a mail filter for EXMH that uses the Naive Bayes classifier to learn classification patters of the user without additional communication between ifile and the user (beyond what is already provided by EXMH). During my Junior year at Carnegie Mellon, I worked under the supervision of Tom Mitchell in order to formalize my work and produce some scientific evidence of its worthiness. Read the paper.

• WordNet
  WordNet is a semantic network created by the Princeton CogSci department that provides an incredible amount of information about a large portion of the English language. Of course it is neither all-encompassing nor flawless, but one could imagine that it could act as beneficial background knowledge in text-related tasks. I'm currently exploring the possibility of using WordNet in tasks such as data mining and classification.

  I wrote a Perl module to allow easy access WordNet synonyms, definitions, hypernyms, hyponyms, meronyms and holonyms. It requires the WordNet package.
  Dan Brian is working on a nice Web interface to WordNet.

• Efficient Web Spidering with Reinforcement Learning
  While at Just Research, I worked on the problem of intelligently spidering the Web for documents of a certain type or topic. In general, directing a spider according to the text found on web pages can significantly improve the rate at which papers are found. In certain domains, a reinforcement learning framework can boost that performance even further to achieve very efficient on-topic page retrievals. Read the paper.

• Naive Bayes
  Naive Bayes is a text classification algorithm that is surprisingly effective for its simplicity. You will find mentionings of it in much of my work. I generally use Andrew McCallum's Bow package to run experiments that involve naive Bayes.

  A Comparison of Event Models for Naive Bayes Text Classification (Andrew McCallum and Kamal Nigam - published in AAAI-98 Workshop on "Learning for Text Categorization") provides a nice introduction to the algorithm, related issues and datasets commonly used to evaluate text classification algorithms.

Publications

• Jason Rennie and Andrew McCallum. 1999. Using Reinforcement Learning to Spider the Web Efficiently. Proceedings of the Sixteenth International Conference on Machine Learning (ICML-99). (Workshop version with new results also available)
• Andrew McCallum, Kamal Nigam, Jason Rennie and Kristie Seymore. 1999. Building Domain-Specific Search Engines with Machine Learning Techniques. AAAI-99 Spring Symposium.
• Jason Rennie. 1998. ifile: An Application of Machine Learning to Mail Filtering.

Machine Learning/Artificial Intelligence

• General: People, Resources, CoraSearch, NgSearch, Web IR & IE,
• ML Conferences: ICML (Text), NIPS, AAAI, JADT(SATD) SIGIR, Agents, IJCAI, ACL, ANLP2000, IUI, IAAI, ECML, UAI, ICTAI, SCIE, TREC,
• KDD Conferences: PAKDD, DaWaK, ISADS, IFIP,
• Other Conferences: USENIX, WWW,
• Journals: DM/KD, JAIR, IDA, JCIF, Instance Selection (DMKD),
• Groups: AAAI, ACM, Brown, MIT, CMU Text, CMU RL, UCI ML Repository, MIT SCM, Auton Lab, Topics: Rule Extraction,
• Companies: MITRE, WiseWire, JustResearch,
• Data: UCI ML, UCI KDD, 20 Newsgroups, Reuters-21578, WebKB, Yahoo!,
• Tools: Standards (local), Stemming, (Perl), Lexing, Libbow, ifile, Decision Trees, WordNet (Perl code), Link Grammar, SVMs,
• Interesting Papers

Resources

• MITRE: Info, Home, AI-Contrib, Foo, CVW,
• Boston: Boston Forecast, Bedford Forecast, Yahoo!, Doppler (Regional), Cambridge, Massachusetts,
• Perl: perl.org, perl.com, perl5, CPAN, FAQ, WordNet,
• Linux: LinuxHQ, Kernel, SlashDot, FreshMeat, Org, HowTo, RedHat, Debian, Stampede, tytso, SunSite, Games, Software, Apps, ssh (ftp), Java, FVWM (IconMan), Mouse Wheel,
• Unix: Unix, CVS (Download), PRCS, MH, EXMH, SunSite, Make, GIMP, LaTeX, 2HTML, BibTeX, XYPic, DB, TTH (manual), gdb, GNUPlot (manual),
• WWW: HTML, Java, mod_perl, CGI, TTH, CSS,
• Search: Google, Maps, AltaVista, Yahoo, Lycos, HotBot, Patents, Area Codes, List, Search, Auctions,
• Car: Blue Book, NADA, AAA, Consumer Reports, CarPoint,
• Pittsburgh: Forecast, Doppler (Regional), Buses, Fencing, Movies,
• Class: Graph Theory, Math History,
• C Code: Standards (local), GTK, gnu.org,
• Hardware: ChipMerchant, CPU Solutions, Memory World, PriceWatch, Emachines,
• Languages: ML, CaML, Java (docs), Python (WordNet),
• XML: XML, XML-QL, Xtract, SGML, SParser, JClark,
• CS Papers: Cora, Tech Reports, XXX LANL, CoRR,
• Travel: Intl Warnings, USAir, Delta, Northwest, United, American,
• Govt: GIDEP, NIMA-GNS, LDC,
• Other: EBay, FSM Applet, Answers, Coke, HUB, Randomized Algorithms, Commencement, Drinks, HospitalWeb,
  
• Fun: Onion, Forum2000, CMU CS,
  * = internal

People

• Undergrads:

  Catherine Carbone, Patrick Doane, Mike Haefele, Ruth Jamison, Iain Keddie, Stan Kwok, Helen Malyutin, Shannon Qiu, Trey Smith, Aaron Siegel, Chris Tchou, Robert Watson,

• Grads:

  Doug Baker, Don Blaheta, Justin Boyan, Carl Burch, Rob Cooley, Dayne Freitag, Thorsten Joachims, Rosie Jones, Andrew Ng, Kamal Nigam, Joseph O'Sullivan, Chuck Rosenberg, Kristie Seymore, Sean Slattery,

• Researchers:

  Shumeet Baluja, Soumen Chakrabarti, Mark Craven, Thomas Hofmann, Andrew McCallum, Marina Melia, Mehran Sahami, Rakesh Agrawal David Lewis,

• Professors:

  Avrim Blum, Rich Caruana, Eugene Charniak, Oren Etzioni, Ken Haase, David Haussler, Tommi Jaakola, Michael Jordan, Leslie Kaelbling, David Karger, Philip Klein, Peter Lee, David MacKay, Tom Mitchell, Andrew Moore, Michael Pazzani, Roni Rosenfeld, Danny Sleator, Paul Viola, Yiming Yang,

Interests

• Hacking/Coding
• Sports: Tennis, Bowling, Table Tennis, Fencing, Racquetball, etc.
• Computer Games: Diablo, Starcraft, SimCity 2000, etc.
• Books: Sci-Fi, Fantasy, Science, Philosophy
• Music: Rush, Genesis, Steve Hackett, King Crimson, Yes, No
• Boots

Other

• 1997 PGSS Presentation
• List of classes I have taken.

Geek Code

-----BEGIN GEEK CODE BLOCK-----
Version: 3.1.2
GCS/M d(-) s+:- a-- C++$ UL+++$ P+++$ L+++$ E(+) W++ N+ o? K?
w-- O- M(-) V- PS PE+ Y+ PGP>+ t(+) 5 X(+) R+ tv- b+ DI+ D++ G
e++>++++* h--- r++>+++ y?
------END GEEK CODE BLOCK-----