Extraction of Information from Speech and Text: 520.476
Department of Electrical and Computer Engineering
The Johns Hopkins University
Spring 1997

o

Introduction to methods of speech and text understanding that serve as the basis of future person-to-person and person-with-database communication. The constructed systems serve as on- and off-ramps to the information superhighway, using it to address vaguely specified data residing in imprecisely known locations. The course is a natural continuation of 520.475 and/or 600.465 but is independent of either. Topics include elementary information theory, hidden Markov models, efficient hypothesis search methods, statistical decision trees, the expectation-maximization (EM) algorithm, maximum entropy estimation, context-free-grammars, parsing, and the Baum, Viterbi and CYK algorithms.

Suggested in sequence with
ECE 520.475 Processing and Recognition of Speech
CS 600.465 Introduction to Natural Language Processing.
3 credits

Instructor: Prof. Frederick Jelinek, 320 Barton Hall, jelinek@jhu.edu
Office Hours: .

Teaching Assistant: Ciprian Chelba, 320 Barton Hall, chelba@jhu.edu
Office Hours: Mo and Wed, 5-6 PM, 320 Barton Hall.

Course Material: Homeworks, projects, ... This link is only available to .jhu.edu users.

Prerequisites: No formal pre-requisites. Basic familiarity with probability and C/C++ programming are expected.

Lectures: Thursday and Friday, 9:00 - 10:15 AM in Maryland Hall 104.
Make-up classes: Monday, Feb 10; Wed, March 5; Wed, March 26 in Remsen Hall 101.

Text:
EN 520.476 Extraction of Information from Speech and Text --- Course Notes, Spring 1997
Frederick Jelinek
Available from printing services, Garland Hall, basement

Homework and Projects: Homework will be assigned regularly. Homeworks are expected to be entirely the student's own work. Projects are supposed to be worked on in teams of 2-4 people.

Grading: homework
projects
??%
??%
Graduate students will be held to higher standards than undergraduates. Projects are a major part of the course. Students will work in teams to complete the project, but each student will be graded individually.


Last Modified: 1997/02/03, chelba@jhu.edu