CS 838 - Machine Learning and Information Retrieval (Fall 1995)

The class will meet in from 3 - 5:30 on Thurdays; the call number is 73206. The official classroom is 138 Pyschology, however we will meet in 2310 CS & Stats unless someone else needs that room.

Prerequisite: CS 540 (CS 760 would be helpful)

Class mailing list: mlir@cs.wisc.edu (archive of old email)

Schedule of Talks (or here if your browser doesn't support tables)

This page will be continually under construction throughout the semester. Suggestions about papers to read should be sent to the instructors (belew@cs.wisc.edu and shavlik@cs.wisc.edu).

Overview

Requirements

1. Students will be expected to present 2-3 papers.

2. Each paper will be also assigned two `commentators.' Commentators will write a one-page critique of the paper and will lead the discussion on the paper.

3. Students will be expected to read each week's papers and participate in the discussions.

4. Students will do substantial class projects of their choosing. Project proposals will be due early in the semester and a progress report will be due mid-semester. Team projects are acceptable.

Potential Topics

• IR basics
  • Vector model
  • Relevance Feedback
  • Probalistic models
  • Evaluation
  • Corpus-based linguistics

• Defining the IR problem(s)
  • Text Categorization
  • Routing/Filtering
  • Text segmentation
  • Clustering
  • User Preferences/Usage Pattern Analysis
  • Browsing
  • Multi-Source integration

Relevant Research

Background reading

Machine Learning for information retrieval: Neural networks, symbolic learning and genetic algorithms. by H. Chen, JASIS 46(3):194-216, April 1995.

Optional reading

Belew Group's Papers

• Connectionist representations for IR

  Adaptive information retrieval: Using a connectionist representation to retrieve and learn about documents, by R. K. Belew. Proc. SIGIR-89, pp. 11-20, Cambridge, MA, 1989, ACM Press.

  A Symbolic and Connectionist Approach to Legal Information Retrieval, by D. E. Rose, 1994, Hillsdale, NJ: Erlbaum.

• Learning to reduce dimensionality

  Latent Semantic Indexing is an optimal special case of multidimenional scaling, by B. T. Bartell, G. W. Cottrell, and R. K. Belew, Proc. SIGIR-92, New York, 1992. ACM Press.

  Abstract

• Learning key parameters of IR systems

  Automatic combination of multiple ranked retrieval systems, by B. T. Bartell, G. W. Cottrell, and R. K. Belew. Proc. SIGIR-94, Dublin, 1994, ACM Press.

  Abstract

  Learning the optimal parameters in a ranked retrieval system using multi-query relevance feedback, by B. T. Bartell, G. W. Cottrell, and R. K. Belew. Proc. Symp. on Document Analysis and Information Retrieval, Las Vegas, 1994.

  Abstract

• Statistical patterns in language use

  Exporting phrases: A statistical analysis of topical language, by A. M. Steier and R. K. Belew, Proc. 2nd Symp. on Document Analysis and Information Retrieval, pp. 179-190, 1993.

  Abstract

  Talking about AI: Socially-defined linguistic subcontexts in AI, by A. M. Steier and R. K. Belew, Proc. AAAI-94, pp. 715-720, Seattle, 1994, AAAI Press.

  Abstract

• Artificial Life applied to Web agents

  Artificial life applied to adaptive information agents, by F. Menczer, R. K. Belew, and W. Willuhn. Proc. 1995 AAAI Spring Symp. on Information Gathering from Heterogeneous, Distributed Environments, Stanford, March 1995, AAAI Press.

  Abstract: Proposes a model, inspired by recent artificial life theory, applied to the problem of retrieving information from a large, distributed collection of documents such as the World Wide Web.

Etzioni Group's Papers

• Category translation: Learning to understand information on the Internet, by M. Perkowitz and O. Etzioni, Proc. IJCAI-95, Montreal, Morgan Kaufmann.

Hammond Group's Papers

• FAQ Finder: A case-based approach to knowledge navigation by K. Hammond, R. Burke, C. Martin, and S. Lytinen, Proc. 1995 AAAI Spring Symp. on Information Gathering from Heterogeneous, Distributed Environments, Stanford, March 1995, AAAI Press.

Holte Group's Papers

• A Learning Agent that Assists the Browsing of Software Libraries. C. Drummond, D. Ionescu and R. Holte (1995). Computer Science Dept. Technical Report TR-95-12 (51 pages).

• A Learning Apprentice For Browsing R. Holte and C. Drummond (1994). AAAI Spring Symposium on Software Agents.

• Accelerating Browsing by Automatically Inferring a User's Search Goal. C. Drummond, R. Holte and D. Ionescu (1993). Proceedings of the Eighth Knowledge-Based Software Engineering Conference, pp. 160-167.

Lashkari's WebHound

• Collaborative interface agents (postscript), by Y. Lashkari, M. Metral, and P. Maes, Proc. 1994 AAAI Conf., Seattle, AAAI Press.

  Abstract: (This isn't a paper on WebHound, but it is the closest published article.)

Lang's News Weeder

• NewsWeeder: Learning to filter NetNews, Proc. 1995 Machine Learning Conf., Lake Tahoe, CA, Morgan Kaufmann.

  Abstract: Describes learning experiments done using NewsWeeder, comparing a tf-idf technique with a Minimum Description Length (MDL) approach.

Mitchell Group's WebWatcher Project

• WebWatcher: A learning apprentice for the World Wide Web, by R. Armstrong, D. Freitag, T. Joachims, and T. Mitchell, Proc. 1995 AAAI Spring Symp. on Information Gathering from Heterogeneous, Distributed Environments, Stanford, March 1995, AAAI Press.

  Abstract: A description of WebWatcher and a comparison of different machine learning approaches to suggest hyperlinks.

• WebWatcher: Machine learning and hypertext, to appear in Fachgruppentreffen Maschinelles Lernen, Dortmund, Germany, August 1995.

  Abstract: A description of how hypertext structure can be used to cluster WWW pages.

Pazzani Group's WebLearner (scroll to bottom of page)

• Learning from hotlists and coldlists: Towards a WWW information filtering and seeking agent, by M. Pazzani, L. Nguyen, and S. Mantik, Proc. AI Tools Conf., Washington, DC, 1995.

  Abstract:A hot and a cold list provide the training examples to a learning algorithm.

Shoham Group's Papers

• An adaptive agent for automated web browsing, by M. Balabanovic, Y. Shoham, and Y. Yun. To appear in the special issue on digital libraries, Journal of Visual Communication and Image Representation 6(4), 1995.

  Abstract

• Learning information retrieval agents: Experiments with automated Web browsing, by M. Balabanovic and Y. Shoham, Proc. 1995 AAAI Spring Symp. on Information Gathering from Heterogeneous, Distributed Environments, Stanford, March 1995, AAAI Press.

  Abstract

Voorhees Group's Papers

• Learning collection fusion strategies for information retrieval, by G. Towell, E. M. Voorhees, N. K. Gupta, and B. Johnson-Laird, Proc. ML-95, pp. 540-548, Lake Tahoe, 1995, Morgan Kaufmann.

• Learning collection fusion strategies, by E. M. Voorhees, N. K. Gupta, and B. Johnson-Laird, Proc. SIGIR '95, pp. 172-179, Seattle, 1995, ACM Press.

Xerox PARC papers

• A neural network approach to topic spotting, by E. Wiener, J. O. Pedersen and A. S. Weigend, Proc. SDAIR '95, pp. 317-332, Las Vegas, NV, 1995.

• A trainable document summarizer, by J. Kupiec, J. Pederson, and F. Chen, Proc. SIGIR '95, pp. 68-73, Seattle, 1995, ACM Press.

• A comparison of classifiers and document representations for the routing problem, by H. Schutze, D. A. Hull, and J. O. Pedersen. Proc. SIGIR '95, pp. 229-237, Seattle, 1995, ACM Press.

Other Papers

• A sequential algorithm for training text classifiers, by D. D. Lewis, W. A. Gale. Proc. SIGIR '94, pp. 3-12. Dublin, Ireland. 1994. Springer-Verlag.

• Learning document category descriptions through the extraction of semantically significant phrases, by B. Krulwich, Proc. IJCAI-95 workshop on Data Engineering for Inductive Learning, 1995.

• Learning user interests across heterogeneous document databases, by B. Krulwich, Proc. 1995 AAAI Spring Symp. on Information Gathering from Heterogeneous, Distributed Environments, Stanford, March 1995, AAAI Press.

• Clustering full text documents, by J. Martin, Proc. IJCAI-95 workshop on Data Engineering for Inductive Learning, 1995.

• Text categorization and relational learning, by W. W. Cohen, Proc. Machine Learning '95 Conf., pp. 124-132, Lake Tahoe, CA, Morgan Kaufmann, 1995.

• An Evolutionary Connectionist Approach to Personal Information Filtering, Michael McElligott and Humphrey Sorensen. Proc. Fourth Irish Neural Network Conference. pp. 141-146. University College, Dublin. Sept. 1994.

• Word Sense Disambiguation Using Statistical Models of Roget's Categories Trained on Large Corpora, by David Yarowsky. Proc. COLING-92. Nantes, Aug 23-28, 1992, pp 454-460.

• Discrimination Decisions for 100,000-Dimensional Spaces, by William A Gale, Kenneth W Church, David Yarowsky, Statistical Research Reports, No. 103, Jan 16, 1992, AT&T Bell Laboratories.

Relevant On-Line Resources

• Invited ML95 talk on Machine Learning and Information Retrieval, by B. Croft.
• A brief bibliography and resource list on NLP for IR, by D. D. Lewis and E. D. Liddy.
• Large-Scale Persistent Object Systems for Corpus Linguistics and Information Retrieval by R. Futrelle and X. Zhang
• Interactive Information Group at the Canadian NRC
• Association for Computational Linguistics (ACL)
• Computational linguistics and NLP e-prints
• Word Net
• Information Filtering Resources
• Intellectual Property and the National Information Infrastructure
• ONR's `Information Finding and Displaying' Project

Other resources relevant to ML generally

• People in ML
• WWW Pages Relevant to the UW ML Research Group
• Recent Table of Contents and Abstracts of Selected ML Journals
  (mostly from Wendt Library; readable only if @cs.wisc.edu)