[Paper Review] Separating Dependency from Constituency in a Tree Rewriting System

Anoop Sarkar|ArXiv.org|Sep 18, 1998

Natural Language Processing Techniques14 references3 citations

TL;DR

This paper introduces Link-Sharing Tree Adjoining Grammar (LSTAG), a novel tree-rewriting formalism that explicitly separates linguistic dependency from constituency in syntactic analysis. By decoupling these two structural aspects, LSTAG enables a more precise and formal representation of coordination phenomena, offering a significant improvement over prior systems that conflated dependency and constituency in tree-rewriting frameworks.

ABSTRACT

In this paper we present a new tree-rewriting formalism called Link-Sharing Tree Adjoining Grammar (LSTAG) which is a variant of synchronous TAGs. Using LSTAG we define an approach towards coordination where linguistic dependency is distinguished from the notion of constituency. Such an approach towards coordination that explicitly distinguishes dependencies from constituency gives a better formal understanding of its representation when compared to previous approaches that use tree-rewriting systems which conflate the two issues.

Motivation & Objective

• To address the limitations of existing tree-rewriting systems that conflate dependency and constituency in syntactic representations.
• To provide a formal framework that explicitly models linguistic dependencies apart from constituent structure.
• To improve the representation of coordination in syntactic structures by distinguishing between dependency and constituency.
• To develop a formalism capable of capturing complex syntactic phenomena with greater precision than traditional synchronous TAGs.
• To lay the foundation for more accurate syntactic parsing and linguistic analysis by decoupling structural components.

Proposed method

• Proposes Link-Sharing Tree Adjoining Grammar (LSTAG), a variant of synchronous Tree Adjoining Grammars (TAGs).
• Uses a dual-structure approach where dependency and constituent structures are maintained separately but linked via shared links.
• Employs tree-rewriting rules that generate and combine syntactic structures while preserving distinct dependency and constituent relations.
• Introduces a formal mechanism for sharing links between dependency and constituent trees to maintain structural coherence.
• Applies the formalism to model coordination, ensuring that dependency relations are not conflated with constituent boundaries.
• Uses a formal grammar framework that supports both constituent structure and dependency structure as first-class components.

Experimental results

Research questions

• RQ1 How can dependency and constituency be formally separated in a tree-rewriting system?
• RQ2 What formalism allows for precise representation of coordination without conflating dependency and constituent structure?
• RQ3 Can a tree-rewriting system maintain both dependency and constituent structure independently while ensuring coherence?
• RQ4 How does decoupling dependency and constituency improve the modeling of syntactic phenomena like coordination?
• RQ5 What advantages does LSTAG offer over traditional synchronous TAGs in syntactic representation?

Key findings

• LSTAG successfully separates dependency and constituency in syntactic representations, enabling clearer modeling of syntactic phenomena.
• The formalism allows for more precise analysis of coordination by decoupling dependency relations from constituent boundaries.
• By maintaining distinct structures linked via shared links, LSTAG avoids the ambiguity introduced by conflating dependency and constituency.
• The approach provides a more formal and transparent understanding of syntactic structure compared to prior systems.
• The framework demonstrates feasibility in modeling complex syntactic constructions with improved structural clarity.
• The results suggest that separating dependency and constituency leads to a more accurate and linguistically plausible syntactic representation.