Constituency Parsing as an Instance of the M-monoid Parsing Problem

TL;DR

This paper models constituency parsing as an instance of the M-monoid parsing problem, enabling the use of a generic algorithm to efficiently compute all recognized constituent trees for a given string.

Contribution

It demonstrates that constituency parsing can be formulated within the M-monoid framework and applies a generic parsing algorithm to solve it for a class of constituent tree automata.

Findings

The constituency parsing problem is an instance of the M-monoid parsing problem.

A generic M-monoid parsing algorithm can be employed for constituency parsing.

The approach applies to a meaningful class of constituent tree automata.

Abstract

We consider the constituent parsing problem which states: given a final state normalized constituent tree automaton (CTA) and a string, compute the set of all constituent trees that are inductively recognized by the CTA and yield the string. We show that this problem is an instance of the M-monoid parsing problem. Moreover, we show that we can employ the generic M-monoid parsing algorithm to solve the constituency parsing problem for a meaningful class of CTA.