Efficient probabilistic top-down and left-corner parsing

TL;DR

This paper explores efficient top-down and left-corner probabilistic parsing methods that avoid dynamic programming, demonstrating that incorporating non-local information enhances both accuracy and efficiency in broad-coverage parsing.

Contribution

It introduces and compares top-down and left-corner parsing approaches, highlighting the benefits of non-local information for improved accuracy and search efficiency.

Findings

Both parsing methods are viable alternatives to bottom-up parsing.

Non-local information improves parser accuracy.

Non-local information significantly enhances search efficiency.

Abstract

This paper examines efficient predictive broad-coverage parsing without dynamic programming. In contrast to bottom-up methods, depth-first top-down parsing produces partial parses that are fully connected trees spanning the entire left context, from which any kind of non-local dependency or partial semantic interpretation can in principle be read. We contrast two predictive parsing approaches, top-down and left-corner parsing, and find both to be viable. In addition, we find that enhancement with non-local information not only improves parser accuracy, but also substantially improves the search efficiency.