You are here: Wagsoft · Papers

Sentence Analysis and Generation
-- a Systemic Perspective

Michael O'Donnell

Ph.D. Dissertation,
Linguistics Dept.,
University of Sydney.


The thesis describes a computational system for the analysis and generation of sentences using Systemic-Functional Linguistics (SFL). It represents the first description of a bi-directional system using Systemic grammar. Bi-directional systems are rare regardless of the formalism used, and are usually restricted to grammatical processing only, while the present system analyses to and generates from a semantic representation. The thesis is structured in two parts:

Part A discusses the Systemic resources used for linguistic representations (the 'grammar' of a language). The sentence is modeled tri-stratally: in terms of semantics (ideational, interactional and textual semantics), lexico-grammar, and graphology. The mapping between these strata is also described. A Systemic-Functional framework is used.

Part B describes the processes which use these resources, focusing on single-sentence analysis and generation. Theoretical issues in sentence processing are raised, with particular focus on their application to processing with a Systemic grammar.

Novel contribution has also been made in several specific areas, particularly in regards to Systemic parsing and generation on the process side, and inter-stratal mapping on the resource side. These contributions are discussed in the relevant sections.

The discussion is based on my implementation of a sentence analysis and generation system, called the WAG system -- Workbench for Analysis and Generation.

Chapters (for download)

Chapter 1 Introduction

1. Problems of Integration in Modeling
2. Single Sentence Representation and Processing
3. Process & Resource
4. Declarativisation: Process-Resource Separation
5. The WAG System
6. Contributions of this Work
7. Thesis Structure

Part A Resource Model

Chapter 2 The Micro-Resource Model

1. Introduction
2. Components of the Resource Model
3. The Systemic Formalism
4. Lexico-Grammatical Resources
5. Graphological Resources
6. The Lexicon
7. Summary of Micro-Resource Model

Chapter 3 Ideational Representation

1. Ideational Representation
2. Generalised Ideational Models
3. Institution-Specific Ideational Models
4. A Combined Approach
5. Multiple Domains in a Single Text
6. Summary of Ideational Representation

Chapter 4 Interactional Representation

1. Micro-Representation of Interaction
2. Move Structure
3. The Speech-Act Network
4. Negotiatory Moves
5. The Object of Negotiation: Information or Action
6. The Object of Elicitation: The ‘Required’ element
7. Summary

Chapter 5 Textual Resource

1. The Nature of Textual Systems
2. Themacity
3. Rhetorical Relevance
4. Information Status
5. Summary of Textual Resources

Chapter 6 Mapping Semantics and Lexico-grammar

1. What is Inter-Stratal Mapping?
2. The Chooser-Inquiry Interface
3. WAG: Feature Selection Constraints
4. Some Descriptions
5. Some Extensions
6. Summary of Inter-stratal Mapping

Part B Process Model

Chapter 7 Process Model Overview

1. Process Model Overview
2. Some Processing Issues
3. Processing Demands on Resources
4. Summary

Chapter 8 Systemic Unification & Feature Logic

1. Introduction
2. Representation and Processing of Systemic Features
3. Representation and Processing of Systemic Structures
4. WAG’s Constraint Language
5. Summary

Chapter 9 Parsing Strategies

1. Introduction
2. Parsing Control Strategies
3. Dealing with Non-Determinism
4. Summary

Chapter 10 The WAG Analyser

1. The Stages of Sentence Analysis
2. Graphological Analysis
3. Lexical Analysis
4. Preparation of the Parsing Grammar
5. Lexico-grammatical Analysis
6. Micro-Semantic Analysis
7. Summary & Conclusions

Chapter 11 Sentence Generation

1. What is Sentence Generation?
2. Some Issues in Sentence Generation
3. WAG’s Input Specification
4. Stages in Systemic Sentence Generation
5. Comparison With Penman
6. Conclusions

Chapter 12 Summary & Conclusions

1. Summary
2. Contributions
3. Conclusions