Topics

Information Retrieval (IR)

• Text query operators and languages.
• Inverted indices, basic data and file structures; fielded queries; structured vs. unstructured queries.
• Relevance ranking: query-to-document proximity
• Basic ranking heuristics: word counts, TF/IDF, term positions, highlighting.
• Vector-space models and document-to-document similarity.
• Document clustering: algorithms and applications.  Inducing topic hierarchies.
• Dimensionality reduction and projection.
• Document classification: rule-driven filtering, Bayesian, SVM.  Near-neighbor computations.
• Low-level "natural language" - query parsing, question-answer systems, summarization.
• Adaptive ranking and personalization - user profiles, rule-driven.  Recommendation systems.
• Information retrieval in hyperlinked environments - ranking, topic distillation, clustering and classification.
• Information retrieval in business environments - load and sizing metrics, access control, interoperability and integration.

• Distributed database design - data fragmentation (horizontal and vertical) and allocation
• Query processing in distributed databases - decomposition, localization, techniques for parallel/distributed sort and join
• Reliable distributed database management - failure models, distributed commit (2PC and 3PC), data replication (RAWA, ROWA, primary copy schemes, available copies scheme, 1 copy serializability), network partitions (quorums and coteries)

Readings: Books and Notes

The basic reading material for this course will be in the form of lecture notes that will be handed out to the students at the beginning of each lecture. Electronic copies of the notes will also be available from the Lectures page. At the end of each lecture, we will provide a list of additional readings relevant to the material covered in that lecture. Pointers to these additional readings will also be listed in the Lectures page. In addition, for the information retrieval component of the course, we recommend that the students consult one or more of the following two textbooks:

• Managing Gigabytes: Compressing and Indexing Documents and Images (2nd Edition)
  Ian Witten, Alistair Moffat, and Timothy C. Bell, Morgan Kauffman, 1999.
  ISBN 1-55860-570-3.
  Home page: www.cs.mu.oz.au/mg/
  
  
• Modern Information Retrieval
  Ricardo Baeza-Yates and Berthier Ribeiro-Neto, Addison Wesley Longman, 1999.
  ISBN 0-201-39829-X.
  Home page: http://www.sims.berkeley.edu/~hearst/irbook/
  
  

Prerequisites

We expect the student to have a sound grasp of basic data structures and computer architecture as well as a mathematical background that includes elementary algebra and probability. In addition, for the distributed databases portion of the course, we expect the student to have taken CS145 (Introduction to Databases) and CS245 (Database System Principles), or at least be familiar with the material covered in those two courses. For the programming project, we will assume prior knowledge of Unix and C programming experience.

Time and Place

The class meets twice a week, on Mondays and Wednesdays, from 12:50-2:50PM, in Thornton 102. On specified days, sections will be conducted by the Projects TA to introduce the VDK software (see Project) and aid in selecting project topics. In addition, review sessions will be held before the midterm and finals. Announcements for these sessions will appear on the class Web page and on the class mailing list.

Communication

Besides office hours, email will be the primary form of communication between the students and the course staff. Doubts and clarifications may be sent to the TAs and the instructor at the email addresses listed in the Personnel page. In addition, a course mailing list

cs347-spr01@lists.stanford.edu

is being maintained. This list will be used mainly by the staff, to send announcements, and convey important and timely information to all the students. We require all students to subscribe to the class mailing list. To sign up for the list, send email to majordomo@lists.stanford.edu from your preferred mail host, with "subscribe cs347-spr01" as the body of the message.

A class newsgroup (su.class.cs347) has been set up for use by the students. This should be used as a discussion forum, for clarifying doubts, and to find project partners. Important announcements will also be posted to this newsgroup. However, the course staff will not be continuously monitoring the traffic on this newsgroup. Therefore, for timely response, questions to the staff should be sent to their individual email addresses.

Group Project

All students will be expected to participate in a group project, involving sets of 4 or 5 students. Each group must build an interesting information retrieval application, demonstrate it in operation, and write a brief report on the implementation. See the Project page for details on project topics, support code, and submission deadlines.

Grading Policy

Grades will depend on the performance in the mid-term and final exams and the group programming project. The individual weights for these components will be as follows:

• Midterm - 30%
• Final - 40%
• Programming project - 30%