[Paper Review] Feature-Based TAG in place of multi-component adjunction: Computational Implications

Beth Ann Hockey, Srinivas Bangalore|ArXiv.org|Oct 26, 1994

Manufacturing Process and Optimization3 citations

TL;DR

This paper proposes using feature-based Tree Adjoining Grammar (TAG) to replace multi-component adjunction in linguistic analyses, enabling existing unification-based Earley-style TAG parsers to handle complex constructions without requiring specialized multi-component parsing. The approach maintains full linguistic coverage for English while simplifying computational requirements.

ABSTRACT

Using feature-based Tree Adjoining Grammar (TAG), this paper presents linguistically motivated analyses of constructions claimed to require multi-component adjunction. These feature-based TAG analyses permit parsing of these constructions using an existing unification-based Earley-style TAG parser, thus obviating the need for a multi-component TAG parser without sacrificing linguistic coverage for English.

Motivation & Objective

• To address the computational challenge of parsing constructions requiring multi-component adjunction in Tree Adjoining Grammar (TAG).
• To demonstrate that feature-based TAG can linguistically and computationally replace multi-component adjunction for English constructions.
• To enable parsing of complex syntactic structures using existing unification-based Earley-style TAG parsers.
• To maintain full linguistic coverage of English without extending parser capabilities.
• To reduce the need for specialized multi-component TAG parsers by re-expressing constructions in feature-based TAG.

Proposed method

• Linguistically motivated analysis of constructions previously claimed to require multi-component adjunction.
• Representation of syntactic features in TAG structures to encode complex dependencies.
• Use of unification-based mechanisms to handle feature constraints during parsing.
• Transformation of multi-component adjunction constructions into equivalent feature-based TAG structures.
• Integration of feature-based TAG analyses into an existing unification-based Earley-style TAG parser.
• Validation of parsing success on constructions previously thought to require multi-component adjunction.

Experimental results

Research questions

• RQ1 Can feature-based TAG adequately model constructions that were previously thought to require multi-component adjunction?
• RQ2 Does the use of feature-based TAG allow existing unification-based Earley-style TAG parsers to handle these constructions without modification?
• RQ3 Is full linguistic coverage of English preserved when replacing multi-component adjunction with feature-based TAG?
• RQ4 What are the computational implications of using feature-based TAG instead of multi-component adjunction?
• RQ5 Can complex syntactic dependencies be captured through feature unification rather than structural multi-adjunction?

Key findings

• Feature-based TAG successfully models constructions previously requiring multi-component adjunction in TAG.
• The proposed analyses are parsable using an existing unification-based Earley-style TAG parser, eliminating the need for a multi-component parser.
• Linguistic coverage for English is maintained without sacrificing syntactic accuracy.
• The computational cost of parsing is reduced by avoiding the complexity of multi-component adjunction.
• The approach demonstrates that feature unification can replace structural multi-adjunction in capturing syntactic dependencies.
• The method enables a uniform parsing architecture for complex constructions without extending parser capabilities.