Introduction to Natural Language Processing (600.465)Feature-Based Tagging

The Task, Again

Feature Selection Problems

Feature-based Tagging

Feature Weight (Lambda) Approximation

Estimation

What we got

The Reality

Search for Features

Adding Features in Blocks, Avoiding the Search for the Best

Types of Features

Ambiguity Classes (AC)

Subtags

Combining Subtags

Smoothing

Overtraining

Results