An Empirical Evaluation of Probabilistic Lexicalized Tree Insertion Grammars

TL;DR

This paper empirically evaluates Probabilistic Lexicalized Tree Insertion Grammars (PLTIG), demonstrating their effectiveness in natural language processing tasks by combining advantages of N-grams and PCFGs, with faster training convergence.

Contribution

It provides the first empirical comparison of PLTIGs with N-grams and PCFGs, highlighting their balanced performance and faster training convergence.

Findings

PLTIGs have language modeling performance comparable to N-grams.

PLTIGs outperform non-lexicalized PCFGs in parsing.

Training of PLTIGs converges faster than PCFGs.

Abstract

We present an empirical study of the applicability of Probabilistic Lexicalized Tree Insertion Grammars (PLTIG), a lexicalized counterpart to Probabilistic Context-Free Grammars (PCFG), to problems in stochastic natural-language processing. Comparing the performance of PLTIGs with non-hierarchical N-gram models and PCFGs, we show that PLTIG combines the best aspects of both, with language modeling capability comparable to N-grams, and improved parsing performance over its non-lexicalized counterpart. Furthermore, training of PLTIGs displays faster convergence than PCFGs.