Optimizing for Measure of Performance in Max-Margin Parsing

TL;DR

This paper presents a method to optimize F1-score directly in max-margin structured SVMs for constituency parsing, improving performance by integrating the specific loss function into the training process.

Contribution

It introduces a novel approach to optimize for F1-score directly in max-margin parsing, focusing on original tree structures rather than binarized ones.

Findings

Improved F1-score in constituency parsing

Effective integration of loss function into structured SVMs

Optimization with respect to original parse trees

Abstract

Many statistical learning problems in the area of natural language processing including sequence tagging, sequence segmentation and syntactic parsing has been successfully approached by means of structured prediction methods. An appealing property of the corresponding discriminative learning algorithms is their ability to integrate the loss function of interest directly into the optimization process, which potentially can increase the resulting performance accuracy. Here, we demonstrate on the example of constituency parsing how to optimize for F1-score in the max-margin framework of structural SVM. In particular, the optimization is with respect to the original (not binarized) trees.