Introduction to Natural Language Processing (600.465)Mutual Information and Word Classes

Table of Contents

Introduction to Natural Language Processing (600.465)Mutual Information and Word Classes

The Problem

Word Classes

Solution

The New Model

Training Data

Training the New Model

Classes: How To Get Them

Creating the Word-to-Class Map

Simplifying the Objective Function

Maximizing Mutual Information(dependent on the mapping r)

Training or Heldout?

The Greedy Algorithm

Word Classes in Applications

Email: hajic@cs.jhu.edu

Home Page: http://www.cs.jhu.edu/~hajic/courses/cs465/syllabus.html