Lynx: Building a Statistical Parser from a Rule-Based Parser

Peter Venable

Abstract

Not only can we derive a probabilistic parser from an existing rule-based parser, but in so doing we can achieve higher accuracy, robustness, and adaptability. In this talk I will give a brief overview of the Link grammar formalism, and then describe Lynx, my new generative grammar model based on Link. Discussing the trade-off between speed and accuracy, I'll show how to parse sentences using this (or any) generative model, finishing up with some experimental results showing how Lynx not only captures the Link grammar model, but appropriatly generalizes it, resulting in improved accuracy and robustness.

Note that this talk is in partial fulfillment of the speaking requirement.


Charles Rosenberg
Last modified: Thu Jun 28 14:05:51 EDT 2001