[Paper Review] Application-driven automatic subgrammar extraction

Renate Henschel, John Bateman|arXiv (Cornell University)|Nov 19, 1997

Natural Language Processing Techniques9 references3 citations

TL;DR

This paper presents an application-driven method for automatically extracting consistent, task-specific subgrammars from large-scale systemic grammars, leveraging formal equivalence between systemic and typed unification grammars. The approach reduces computational overhead while enabling reuse, with evaluation showing effective generation of encyclopedia entries using extracted subgrammars.

ABSTRACT

The space and run-time requirements of broad coverage grammars appear for many applications unreasonably large in relation to the relative simplicity of the task at hand. On the other hand, handcrafted development of application-dependent grammars is in danger of duplicating work which is then difficult to re-use in other contexts of application. To overcome this problem, we present in this paper a procedure for the automatic extraction of application-tuned consistent subgrammars from proved large-scale generation grammars. The procedure has been implemented for large-scale systemic grammars and builds on the formal equivalence between systemic grammars and typed unification based grammars. Its evaluation for the generation of encyclopedia entries is described, and directions of future development, applicability, and extensions are discussed.

Motivation & Objective

• To address the high space and runtime costs of broad-coverage grammars in simple language generation tasks.
• To overcome the redundancy and low reusability of handcrafted application-specific grammars.
• To develop an automatic procedure that extracts consistent, application-tuned subgrammars from proven large-scale generation grammars.
• To ensure the extracted subgrammars maintain formal consistency and linguistic adequacy for target applications.
• To demonstrate feasibility and effectiveness through application to encyclopedia entry generation.

Proposed method

• The method exploits the formal equivalence between systemic grammars and typed unification-based grammars to enable systematic transformation.
• It automatically identifies and extracts relevant grammar components based on application-specific input and output requirements.
• The procedure ensures subgrammar consistency by preserving the formal structure and constraints of the original grammar.
• The approach is implemented within a computational framework supporting large-scale grammar processing and subgrammar derivation.
• Evaluation is conducted on a corpus of encyclopedia entries to validate subgrammar quality and generation performance.
• The system supports extensibility and reuse across different linguistic applications through modular subgrammar extraction.

Experimental results

Research questions

• RQ1 Can consistent, application-specific subgrammars be automatically extracted from large-scale systemic grammars?
• RQ2 How can formal equivalence between systemic and typed unification grammars be leveraged to ensure subgrammar consistency?
• RQ3 To what extent does the extracted subgrammar maintain linguistic adequacy for target generation tasks?
• RQ4 How does the performance of the extracted subgrammar compare to full grammars in terms of efficiency and accuracy?
• RQ5 What are the practical implications for reusability and maintenance of grammar components across different applications?

Key findings

• The automatic subgrammar extraction procedure successfully generates consistent, application-tuned grammars from large-scale systemic grammars.
• The extracted subgrammars are sufficient for high-quality generation of encyclopedia entries, demonstrating practical applicability.
• The method reduces computational resource demands while preserving linguistic correctness and structural consistency.
• The formal equivalence between systemic and typed unification grammars enables reliable and systematic subgrammar derivation.
• The approach enables reuse of grammar components across different applications, reducing duplication and maintenance overhead.
• Evaluation results confirm the feasibility and effectiveness of the method in a real-world generation task.