1 Introduction

1.1 The Research Perspective

1.1.1 Differences between NL Generation and NL Understanding

1.1.2 Sharing Knowledge between Generation and Understanding

1.2 The Applications Perspective

1.2.1 Computer as Authoring Aid

1.2.2 Computer as Author

1.2.3 Uses of NLG Technology

1.3 Some Example NLG Systems

1.3.1 WEATHERREPORTER

1.3.2 FOG

1.3.3 IDAS

1.3.4 MODELEXPLAINER

1.3.5 PEBA

1.3.6 STOP

1.4 A Short History of NLG

1.5 The Structure of This Book

1.6 Further Reading

2 National Language Generation in Practice

2.1 Introduction

2.2 When Are NLG Techniques Appropriate?

2.2.1 Text versus Graphics

2.2.2 Natural Language Generation versus Mail Merge

2.2.3 Natural Language Generation versus Human Authoring

2.3 Using a Corpus to Determine User Requirements

2.3.1 Assembling an Initial Corpus of Output Texts

2.3.2 Analysing the Information Content of Corpus Texts

2.3.3 Creating a Target Text Corpus

2.4 Evaluating NLG Systems

2.5 Fielding NLG Systems

2.6 Further Reading

3 The Architecture of a Natural Language Generation System

3.1 Introduction

3.2 The Inputs and Outputs of Natural Language Generation

3.2.1 Language as Goal-Driven Communication

3.2.2 The Inputs to Natural Language Generation

3.2.3 The Output of Natural Language Generation

3.3 An Informal Characterisation of the Architecture

3.3.1 An Overview of the Architecture

3.3.2 Content Determination

3.3.3 Document Structuring

3.3.4 Lexicalisation

3.3.5 Referring Expression Generation

3.3.6 Aggregation

3.3.7 Linguistic Realisation

3.3.8 Structure Realisation

3.4 The Architecture and Its Representations

3.4.1 Broad Structure and Terminology

3.4.2 Messages

3.4.3 The Document Planner

3.4.4 Document Plans

3.4.5 Microplanning

3.4.6 Text Specifications

3.4.7 Phrase Specifications

3.4.8 Surface Realisation

3.5 Other Architectures

3.5.1 Different Representations and Modularisations

3.5.2 Different Architectures:Integrated Systems

3.6 Further Reading

4 Document Planning

4.1 Introduction

4.1.1 What Document Planning Is About

4.1.2 The Inputs and Outputs of Document Planning

4.1.3 A WEATHERREPORTER Example

4.2 Representing Information in the Domain

4.2.1 What's in a Domain Model?

4.2.2 Domain Modelling for WEATHERREPORTER

4.2.3 Implementing Domain Models

4.2.4 Defining Messages

4.2.5 Determining the Degree of Abstraction in Messages

4.2.6 A Methodology for Domain Modelling and Message Definition

4.3 Content Determination

4.3.1 Aspects of Content Determination

4.3.2 Deriving Content Determination Rules

4.3.3 Implementing Content Determination

4.4 Document Structuring

4.4.1 Discourse Relations

4.4.2 Implementation:Schemas

4.4.3 Implementation:Bottom-up Techniques

4.4.4 A Comparison of Approaches

4.4.5 Knowledge Acquisition

4.5 Document Planner Architecture

4.6 Further Reading

5 Microplanning

5.1 Introduction

5.1.1 Why Do We Need Microplanning?

5.1.2 What's Involved in Microplanning?

5.1.3 The Inputs and Outputs of Microplanning

5.1.4 The Architecture of a Microplanner

5.2 Lexicalisation

5.2.1 Simple Lexicalisation

5.2.2 Simple Lexical Choice

5.2.3 Contextual and Pragmatic Influences on Lexical Choice

5.2.4 Expressing Discourse Relations

5.2.5 Fine-Grained Lexicalisation

5.3 Aggregation

5.3.1 Mechanisms for Sentence Formation

5.3.2 Choosing between Possible Aggregations

5.3.3 Order of Presentation

5.3.4 Paragraph Formation

5.4 Generating Referring Expressions

5.4.1 The Nature of the Problem

5.4.2 Forms of Referring Expressions and Their Uses

5.4.3 Requirements for Referring Expression Generation

5.4.4 Generating Pronouns

5.4.5 Generating Subsequent References

5.5 Limitations and Other Approaches

5.6 Further Reading

6 Surface Realisation

6.1 Introduction

6.2 Realising Text Specifications

6.3 Varieties of Phrase Specifications

6.3.1 Skeletal Propositions

6.3.2 Meaning Specifications

6.3.3 Lexicalised Case Frames

6.3.4 Abstract Syntactic Structures

6.3.5 Canned Text

6.3.6 Orthographic Strings

6.3.7 Summary

6.4 KPML

6.4.1 An Overview

6.4.2 The Input to KPML

6.4.3 Using Systemic Grammar for Linguistic Realisation

6.4.4 Summary

6.5 SURGE

6.5.1 An Overview

6.5.2 The Input to SURGE

6.5.3 Functional Unification Grammar

6.5.4 Linguistic Realisation via Unification

6.6 REALPRO

6.6.1 An Overview

6.6.2 The Input to REALPRO

6.6.3 Meaning-Text Theory

6.6.4 How REALPRO Works

6.6.5 Summary

6.7 Choosing a Realiser

6.8 Bidirectional Grammars

6.9 Further Reading

7 Beyond Text Generation

7.1 Introduction

7.2 Typography

7.2.1 The Uses of Typography

7.2.2 Typography in NLG Systems

7.2.3 Implementing Typographic Awareness

7.3 Integrating Text and Graphics

7.3.1 The Automatic Generation of Graphical Objects

7.3.2 Choosing a Medium

7.3.3 Commonalities between Text and Graphics

7.3.4 Implementing Text and Graphics Integration

7.4 Hypertext

7.4.1 Hypertext and Its Uses

7.4.2 Implementing Hypertext-based HLG Systems

7.5 Speech Output

7.5.1 The Benefits of Speech Output

7.5.2 Text-to-Speech Systems

7.5.3 Implementing Concept-to-Speech

7.6 Further Reading

Appendix:NLG Systems Mentioned in This Book

References

Index

1.1 The Macquarie weather summary for February 1995

1.2 Data input for the WEATHERREPORTER system

1.3 FOG input:a weather system over Canada

1.4 Some example forecasts from FOG

1.5 Some example texts from IDAS

1.6 An example MODELEXPLAINER input:An object-oriented class model

1.7 An example description produced by MODELEXPLAINER from the model in Figure 1.6

1.8 A www page generated by PEBA

1.9 A letter generated by the prototype STOP system

2.1 An example human-authored text in the WEATHERREPORTER domain

2.2 An example daily weather record

2.3 An example target text in the WEATHERREPORTER domain

3.1 Modules and tasks

3.2 The weather summary for February 1995

3.3 The weather summary for August 1995

3.4 The weather summary for January 1996

3.5 The structure of the weather summary in Figure 3.2

3.6 Using typography to indicate discourse structure

3.7 An NLG system architecture

3.8 A definition of some message types in WEATHERREPORTER

3.9 A MonthlyTemperatureMsg message

3.10 A MonthlyRainfallMsg message

3.11 A simple document plan representation

3.12 A simple text specification representation

3.13 Some phrase specification representations

3.14 An abstract syntactic representation for The month had some rainy days

3.15 A lexicalised case frame for The month had some rainy days

3.16 Combining canned text and abstract syntactic structures

3.17 Variations in terminology

3.18 A proto-phrase specification representation of the MonthlyRainfallMsg message

4.1 A knowledge base fragment in PEBA

4.2 A daily weather record

4.3 The document plan corresponding to the text in Figure 4.5

4.4 A documentplan

4.5 The weather summary for July 1996

4.6 Example MonthlyTemperatureMsg and RainEventMsg messages from the document plan shown in Figure 4.3

4.7 The correspondence of temperature values to categories

4.8 The definition of TemperatureSpellMsg

4.9 A TemperatureSpellMsg message

4.10 A message defined as a string

4.11 A message specified as a month and a textual attribute

4.12 A corpus-based procedure for domain modelling and message definition

4.13 A corpus-based procedure for identifying content determination rules

4.14 The RST definition of Elaboration

4.15 PEBA's Compare-And-Contrast schema

4.16 A simple set of schemas for WEATHERREPORTER

4.17 A bottom-up discourse structuring algorithm

4.18 Bottom-up construction of a document plan

4.19 Different combinations of document structuring and content determination

5.1 A simple weather summary

5.2 The top-level document plan for the text in Figure 5.1

5.3 The first constituent of the document plan corresponding to the text in Figure 5.1

5.4 The second constituent of the document plan corresponding to the text in Figure 5.1

5.5 A schematic representation of the document plan in Figures 5.2-5.4

5.6 The top-level text specification corresponding to the text in Figure 5.1

5.7 The phrase specification for Sentence1 in Figure 5.6(The month was slightly warmer than average,with the average number of rain days)

5.8 The phrase specification for Sentence2 in Figure 5.6(Heavy rain fell on the 27th and 28th)

5.9 A blackboard architecture for a microplanner

5.10 A simple microplanner architecture

5.11 Simplified microplanning

5.12 A simple template for RainEventMsgs

5.13 A simple message

5.14 The proto-phrase specification produced by applying the template to amessage

5.15 A simple template for RainSpellMessages

5.16 An algorithm for lexicalising spells

5.17 A template proto-phrase specification for on the ith

5.18 A template proto-phrase specification for on the ith and jth

5.19 A decision tree for realising the Contrast relation

5.20 The weather summary for July 1996 without aggregation

5.21 A proto-phrase specification for Heavy rain fell on the 28th

5.22 The result of a simple conjunction(Heavy rainfell on the 27th and 28th)

5.23 The result of a shared-participant conjunction(Heavy rain fell on the 27th and 28th)

5.24 The result of a shared-structure conjunction(Heavy rain fell on the 27th and 28th)

5.25 Some of Scott and de Souza's heuristics for aggregation

5.26 Some example rules from the AECMA Simplified English guide

5.27 A conservative pronoun-generation algorithm

5.28 Distinguishing descriptions

5.29 An algorithm for producing distinguishing descriptions

5.30 Constructing a referring expression

6.1 A simple PEBA text specification

6.2 A surface form with mark-up annotations for the PEBA text

6.3 The PEBA text as displayed by the presentation system

6.4 The logical structure specification in LATEX form

6.5 The logical structure specification in Word RTF form

6.6 A skeletal proposition as an AVM

6.7 A meaning specification

6.8 A lexicalised case frame

6.9 An abstract syntactic structure

6.10 A canned text structure

6.11 An orthographic string structure

6.12 An input to KPML,from which KPML produces March had some rainy days

6.13 An AVM representation of the structure in Figure 6.12

6.14 A more complex SPL expression(The month was cool and dry with the average number of rain days)

6.15 An SPL which exploits a middle model(March had some rainy days)

6.16 The mood system for the English clause

6.17 A chooser for deciding which determiner should be used

6.18 Some realisation statement operators

6.19 Realisation statements in a network

6.20 An input to SURGE,from which SURGE produces March had some rainy days

6.21 An AVM representation of the structure in Figure 6.20

6.22 A more complex input to SURGE(The month was cool anddry with the average number of rain days)

6.23 An FD for a simple sentence(John sells the car)

6.24 A simple FUF grammar

6.25 The result of initial unification of the input FD with the grammar

6.26 The FD in Figure 6.23 after unification with the grammar in Figure 6.24

6.27 The DSyntS for John helps Mary

6.28 Variations of the DSyntS in Figure 6.27

6.29 A DSyntS for March hadsome rainy days

6.30 An AVM representation of the structure in Figure 6.29

6.31 A more complex DSyntS input to REALPRO (The month was cool and dry with the average number of rain days)

6.32 Common relations in MTT's DSyntS

6.33 Some simple REALPRO grammar rules

6.34 Stages of realisation in REALPRO

7.1 An example output from PEBA

7.2 Using typeface variation for emphasis

7.3 Using a labelled list structure to aid information access

7.4 Information expressed typographically as a decision tree

7.5 Information expressed typographically via a table

7.6 Information expressed compactly

7.7 Declarative mark-up annotations

7.8 Physical mark-up annotations

7.9 A tabular presentation of aggregated information

7.10 Multimodal output from the WIP system

7.11 Graphics-only output from the WIP system

7.12 Text-only output from the WIP system

7.13 The document plan for the output in Figure 7.10

7.14 The architecture of a text-to-speech system

7.15 SABLE mark-ups for controlling speech synthesis