Transducers from Rewrite Rules with Backreferences

TL;DR

This paper extends finite-state transducer algorithms for context-sensitive rewrite rules by incorporating limited backreferencing, enabling more elegant and general solutions in natural language processing tasks.

Contribution

It introduces a method to include limited backreferencing in rewrite rules, enhancing the expressiveness and applicability of finite-state transducers in NLP.

Findings

Allows limited backreferencing in rewrite rules

Produces more elegant finite-state transducer constructions

Enhances the generality of rule-based NLP models

Abstract

Context sensitive rewrite rules have been widely used in several areas of natural language processing, including syntax, morphology, phonology and speech processing. Kaplan and Kay, Karttunen, and Mohri & Sproat have given various algorithms to compile such rewrite rules into finite-state transducers. The present paper extends this work by allowing a limited form of backreferencing in such rules. The explicit use of backreferencing leads to more elegant and general solutions.