Probabilistic Parsing Using Left Corner Language Models

TL;DR

This paper presents a new probabilistic left-corner parser that leverages both top-down and bottom-up information, demonstrating improved parsing performance on the Wall Street Journal corpus, though with some limitations due to data characteristics.

Contribution

It introduces a probabilistic left-corner parsing approach and details how to induce grammars from analyzed data, outperforming simple top-down models on a standard benchmark.

Findings

Outperforms top-down probabilistic context-free grammars on WSJ parsing

Grammar induction from Penn Treebank has limitations due to data structure

Left-corner strategy offers advantages in conditioning rule probabilities

Abstract

We introduce a novel parser based on a probabilistic version of a left-corner parser. The left-corner strategy is attractive because rule probabilities can be conditioned on both top-down goals and bottom-up derivations. We develop the underlying theory and explain how a grammar can be induced from analyzed data. We show that the left-corner approach provides an advantage over simple top-down probabilistic context-free grammars in parsing the Wall Street Journal using a grammar induced from the Penn Treebank. We also conclude that the Penn Treebank provides a fairly weak testbed due to the flatness of its bracketings and to the obvious overgeneration and undergeneration of its induced grammar.