## Constraint-based Natural Language Generation: A Survey (2006)

Citations: | 4 - 2 self |

### BibTeX

@TECHREPORT{Piwek06constraint-basednatural,

author = {Paul Piwek and Kees Van Deemter and Walton Hall and Paul Piwek and Kees Van Deemter},

title = {Constraint-based Natural Language Generation: A Survey},

institution = {},

year = {2006}

}

### OpenURL

### Abstract

This paper contains a survey of 25 papers that deal with constraint-based approaches to Natural Language Generation (nlg). Amongst other things, we discuss the nature of the constraints that have been modelled in nlg (stylistic, syntactic, etc.), their type (hard versus soft), the units to which the constraints are applied (words, sentences, rhetorical structure, etc.) and the approaches to solving constraint satisfaction problems in nlg. The survey starts with a brief introduction to the constraint-based view of nlg and a section that aims to clarify the many-faceted notion of a constraint. The aim of this survey is to provide the reader with an overview of constraint-based approaches to nlg; it does not present an exhaustive list of all the literature in this area. What it does aim for, is to give a reasonably representative sample of the various approaches to constraint-based nlg. Table of Contents

### Citations

3382 |
Data Mining - Practical Machine Learning Tools and Techniques with Java Implementations
- Witten, Frank
- 2000
(Show Context)
Citation Context ...n with pipeline systems). To solve ILP problems lp solve is used. 1 For constructing the classifiers a the Naive Bayes algorithm was used as implemented in the Weka machine learning software package (=-=[60]-=-). 7. Approach to addressing constraints in generation Use of Integer Linear Programming (ILP). 8. Single or many constraints? How are constraints combined? Multiple constraints dealt with through ILP... |

1140 |
Integer and Combinatorial Optimization
- Nemhauser, Wolsey
- 1988
(Show Context)
Citation Context ...ot applicable. The constraints are mainly non-linguistic. They are used to formalize the problem of aggregating the outcomes of a set of classifiers in terms of integer linear programming (ILP; e.g., =-=[41]-=-). For example, there is a constraint which requires that each classifier assigns exactly one label. The constraints are on the variables of the ILP target function. This function itself can be seen a... |

1052 |
Consistency in Networks of Relations
- Mackworth
- 1977
(Show Context)
Citation Context ...constraint satisfaction. 8. Single or many constraints? How are constraints combined? Any finite number of constraints (as expressed in a referring expression). Dale and Haddock use A.K. Mackworth’s (=-=[32]-=-) notion of consistency in constraint networks (also applied in Mellish [37]). 9. Any other comments on the paper Essentially the same approach was later used for the generation of (distributive or co... |

854 |
The Bargaining Problem
- Nash
- 1950
(Show Context)
Citation Context ...ms of its satisfaction of the turn and emphasis constraints. The plans are ranked in terms of the product of the scores. Multiplication is proposed as a fair way of comparing scores, following Nash’s =-=[40]-=- work in Game Theory (known as the Nash arbitration plan). 8. Single or many constraints? How are constraints combined? Two soft constraints. See above. 9. Any other comments on the paper – 4.19 Power... |

845 |
Rhetorical structure theory: Toward a functional theory of text organization
- Mann, Thompson
- 1988
(Show Context)
Citation Context ...he constraint(s) (if available) – 1 http://www.geocities.com/lpsolve/ 27s2. Brief informal description of the constraint(s) The goals is build the “best” legal Rhetorical Structure Theory (RST) tree (=-=[33]-=-) for a set of facts and relations between these facts (not all relations need to be included in the tree). Constraints are formalized as scores that are assigned to based on the properties of RST tre... |

681 | Centering: A framework for modelling the local coherence of discourse
- Grosz, Joshi, et al.
- 1996
(Show Context)
Citation Context ...per – 4.9 Kibble and Power [28], 2004, ‘Optimizing Referential Coherence in Text’ 1. Name of the constraint(s) (if available) Constraints are reformulations of well-known rules from Centering Theory (=-=[20]-=-). The named constraints are: cohesion, salience, cheapness and continuity. 2. Brief informal description of the constraint(s) Centering theory is used to inform generation of coherent text and choice... |

417 | Building Natural Language Generation Systems - Reiter, Dale - 2006 |

402 |
Constraint Satisfaction in Logic Programming
- Hentenryck
- 1988
(Show Context)
Citation Context ...d constraints is basically a special case of optimization (which covers both hard and soft constraints). An older and more restrictive notion of constraints can be found in, for example, pages 3–4 of =-=[23]-=-: “Formally, a CSP can be defined in the following way. Assume the existence of a finite set I of variables {X1,X2,...,Xn}, which take respectively their values from their finite domains D1,D2,...,Dn ... |

213 |
Generating Natural Language under Pragmatic Constraints. Lawrence Erlbaum Associates
- Hovy
- 1988
(Show Context)
Citation Context ...nd Robin [15], 1997, ‘Floating Constraints in Lexical Choice’ . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 4.6 Gardent [16], 2002, ‘Generating Minimal Definite Descriptions’ . 20 4.7 Hovy =-=[24, 25]-=-, 1988, ‘Generating Natural Language under Pragmatic Constraints’ . . . . . . . . . . . . . . . . . . . . . . . . . . 21 4.8 Inui, Tokunaga and Tanaka [26], 1992, ‘Text Revision: A Model and Its Imple... |

192 |
P~unning English sentences
- Appelt
- 1984
(Show Context)
Citation Context ... than simple mapping of information and involves addressing multiple constraints at once is certainly not new. One of the first to express it very concisely and work it out in great detail is Appelt (=-=[3]-=-). Although he speaks of ‘goals’ rather than ‘constraints’, his ideas resonate with the approaches we discuss in this survey: “One must constantly bear in mind that language behavior is part of a cohe... |

168 | Pragmatics and natural language generation
- Hovy
- 1990
(Show Context)
Citation Context ...ucted stepwise. The construction can be carried out and controlled in a number of different ways: ∗ Incremental using monitoring regarding the extent to which various constraints are being satisfied (=-=[24, 25]-=-). ∗ Construction using a unification-based grammar (e.g., [15, 44]). ∗ Construction as a sequence of decisions, where alternatives for each decision problem are ranked based on scores which have been... |

158 | Some aspects of optimality in natural language interpretation
- Blutner
- 2000
(Show Context)
Citation Context ...ways satisfy the constraint that is currently least satisfied): Least satisfied precedence [24] Weighted addition [28, 10, 36, 38, 49] Multiplication [48] Bidirectional Superoptimality [57] (based on =-=[8]-=-) 3.5 Approaches to satisfying constraints In this section we organize the papers covered by this survey using a taxonomy. The taxonomy systematizes a variety of approaches for satisfying constraints ... |

129 | Fast Decoding and Optimal Decoding for Machine Translation
- Germann, Jahr, et al.
- 2001
(Show Context)
Citation Context ... directly integrated with a soft constraints that are expressed in terms of a linear function which needs to be maximized/minimized (see section 2.3). This approach has been adopted in [36]. See also =-=[18]-=- for an application of Linear Programming to Machine Translation. • Search using Heuristics – Building a solution A solution is constructed stepwise. The construction can be carried out and controlled... |

93 |
Tree adjoining grammars and lexicalized grammars
- Joshi, Schabes
- 1991
(Show Context)
Citation Context ...ion functions can be parameterized such that users can set coefficients and assign weights. For evolution, they suggest three (non-monotonic) operators: add, delete and change. These operate on LTAG (=-=[27]-=-) derivation trees with a flat semantics. Only candidates that had a high score according to the evaluation function are mutated (in fact, they spawn identical children to which the operators are appl... |

88 | Possibility Theory in Constraint Satisfaction Problems: Handling Priority - Dubois, Fargier, et al. - 1996 |

78 | The practical value of n-grams in generation
- Knight
- 1998
(Show Context)
Citation Context ...son domains for representing word features, and on relationships between features of words, e.g., regulating the position of heads.] 9. Any other comments on the paper None. 4.11 Langkilde and Knight =-=[29, 31]-=-, 1998, ‘Generation that Exploits Corpus-based Statistical Knowledge’ 1. Name of the constraint(s) (if available) – 2. Brief informal description of the constraint(s) See below. 3. Type of constraint(... |

56 | Experiments using stochastic search for text planning
- Mellish, Knott, et al.
- 1998
(Show Context)
Citation Context .... . . . . . 25 4.13 Marciniak & Strube [36], 2005, ‘Beyond the Pipeline: Discrete Optimization in NLP’ . . . . . . . . . . . . . . . . . . . . . . . . 26 4.14 Mellish, Knott, Oberlander and O’Donnell =-=[38]-=-, 1998, ‘Experiments Using Stochastic Search for Text Planning’ . . . . . . . . . 27 4.15 Moriceau and Saint-Dizier [39], 2004, ‘A Constraint-Based Model for Preposition Choice in Natural Language’ . ... |

53 | Organizational structure
- Scott
- 1975
(Show Context)
Citation Context ...constraint applies Text structures. 5. Nature of constraint(s) Discourse level (Centering Theory). 6. Theoretical work or implemented Implemented as an extension of the Iconoclast text planner ([49]; =-=[50]-=-). 7. Approach to addressing constraints in generation The text planner, described in [49], generates a number of text structures. For each text structure alternative realizations in terms of Centerin... |

47 |
Floating constraints in lexical choice
- Robin, McKeown
- 1997
(Show Context)
Citation Context .... . . . . . . . . . . . . . 17 4.4 van Deemter [57], 2004, ‘Towards a Probabilistic Version of Bidirectional OT Syntax and Semantics’ . . . . . . . . . . . . . . . . 18 4.5 Elhadad, McKeown and Robin =-=[15]-=-, 1997, ‘Floating Constraints in Lexical Choice’ . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 4.6 Gardent [16], 2002, ‘Generating Minimal Definite Descriptions’ . 20 4.7 Hovy [24, 25], 198... |

46 |
Notes on connectionism and harmony theory in linguistics
- Prince, Smolensky
- 1991
(Show Context)
Citation Context ...zations for priority levels P and also for the way the priority level of a constraint influences whether it can be dropped. Two important instantiations of ordered constraints — optimality theoretic (=-=[51]-=-) and probabilistic — can be characterized as follows: • Optimality Theory Each constraint c ∈ C is assigned a natural number representing its position in the priority ranking π(c) (lower numbers indi... |

43 | Generating minimal definite descriptions
- Gardent
- 2002
(Show Context)
Citation Context ...sm to make processing of floating constraints efficient. 8. Single or many constraints? How are constraints combined? Multiple constraints. See above. 9. Any other comments on the paper – 4.6 Gardent =-=[16]-=-, 2002, ‘Generating Minimal Definite Descriptions’ 1. Name of the constraints (if available) No names provided. The algorithm focuses on the content determination aspect of Generation of referring Exp... |

40 |
Content determination in the generation of referring expressions
- Dale, Haddock
- 1991
(Show Context)
Citation Context ...eneration Problems’ . . . . . . . . . . . . . . 16 4.2 Callaway and Lester [9], 1997, ‘Dynamically Improving Explanations: A Revision-Based Approach to Explanation Generation’ . 16 4.3 Dale & Haddock =-=[12]-=-, 1991, ‘Content Determination in the Generation of Referring Expressions’ . . . . . . . . . . . . . . . . . . 17 4.4 van Deemter [57], 2004, ‘Towards a Probabilistic Version of Bidirectional OT Synta... |

38 |
Using argumentation to control lexical choice: a unification-based implementation
- Elhadad
- 1993
(Show Context)
Citation Context ...s from conceptual input to lexical items. 19s5. Nature of constraint(s) Constraints on lexical choice. 6. Theoretical work or implemented Implemented in advisor-ii system using the fuf/surge package (=-=[14]-=-). 7. Approach to addressing constraints in generation fuf ([14]) Unification based top-down control regime with dependency-directed mechanism to make processing of floating constraints efficient. 8. ... |

37 |
Computer Interpretation of Natural Language Descriptions
- Mellish
- 1985
(Show Context)
Citation Context ...combined? Any finite number of constraints (as expressed in a referring expression). Dale and Haddock use A.K. Mackworth’s ([32]) notion of consistency in constraint networks (also applied in Mellish =-=[37]-=-). 9. Any other comments on the paper Essentially the same approach was later used for the generation of (distributive or collective) references to sets by Matthew Stone ([56]). Gardent’s ([16]) use o... |

37 | Context modeling and the generation of spoken discourse. Speech Communication 21(1/2):101–121 - Deemter, Odijk - 1997 |

31 | Optimising referential coherence in text generation
- Kibble, Power
- 2004
(Show Context)
Citation Context ... . . . . . . . . . . . . . . . . . . 21 4.8 Inui, Tokunaga and Tanaka [26], 1992, ‘Text Revision: A Model and Its Implementation’ . . . . . . . . . . . . . . . . . . . . . . . 22 4.9 Kibble and Power =-=[28]-=-, 2004, ‘Optimizing Referential Coherence in Text’ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 1s4.10 Langkilde-Geary [30], 2004, ‘An Exploratory Application of Constraint Optim... |

28 |
Writing research and the writer
- Hayes, Flower
- 1986
(Show Context)
Citation Context ...; constraints are used as guards on revision operations rather than goals in themselves. 9. Any other comments on the paper The approach is compared with how human writers produce text. They refer to =-=[22]-=-, who hypothesize that human writers produce multiple drafts to cope with the complexity of writing text. 4.3 Dale & Haddock [12], 1991, ‘Content Determination in the Generation of Referring Expressio... |

27 |
Text Revision: a Model and Its Implementation
- Inui, Tokunaga, et al.
- 1992
(Show Context)
Citation Context ...mal Definite Descriptions’ . 20 4.7 Hovy [24, 25], 1988, ‘Generating Natural Language under Pragmatic Constraints’ . . . . . . . . . . . . . . . . . . . . . . . . . . 21 4.8 Inui, Tokunaga and Tanaka =-=[26]-=-, 1992, ‘Text Revision: A Model and Its Implementation’ . . . . . . . . . . . . . . . . . . . . . . . 22 4.9 Kibble and Power [28], 2004, ‘Optimizing Referential Coherence in Text’ . . . . . . . . . .... |

26 | Planning texts by constraint satisfaction
- Power
- 2000
(Show Context)
Citation Context .... . . . . . . . . . . . . . . . . . . . . . . . 32 4.18 Piwek and Van Deemter [48], 2003, ‘Dialogue as Discourse: Controlling Global Properties of Scripted Dialogue’ . . . . . . . . . . 33 4.19 Power =-=[49]-=-, 2000, ‘Planning text by constraint satisfaction’ . . . . 33 4.20 Reiter [52], 2000, ‘Pipelines and Size Constraints’ . . . . . . . . . 34 4.21 Thiam Chye [10], 2005, ‘Multi-document Summarization by... |

24 |
Guide to constraint programming,” http://kti.ms.mff.cuni.cz/∼bartak/constraints/intro.html
- Bartak
(Show Context)
Citation Context ...t). Note that e does not need to be unique: a CSP can have multiple solutions. Different constraint satisfaction problems can be obtained by varying the components listed above, for example (based on =-=[4]-=-): • Fuzzy Constraint Satisfaction Problem 1. Variables ranging over finite domains. 2. Satisfaction levels are represented as real numbers between 0 and 1. 3. Constraints are combined using min (c1 ⊗... |

24 | Dynamically Improving Explanations: a Revision-Based Approach to Explanation Generation
- Callaway, Lester
- 1997
(Show Context)
Citation Context ...t [6], 2000, ‘Liliput: A Parameterisable FiniteDomain Constraint Solving Framework and Its Evaluation with Natural Language Generation Problems’ . . . . . . . . . . . . . . 16 4.2 Callaway and Lester =-=[9]-=-, 1997, ‘Dynamically Improving Explanations: A Revision-Based Approach to Explanation Generation’ . 16 4.3 Dale & Haddock [12], 1991, ‘Content Determination in the Generation of Referring Expressions’... |

24 |
Generating recipes: An overview of epicure
- Dale
(Show Context)
Citation Context ...y Properties and relations expressed in a referring expression. 5. Nature of the constraints Semantic (i.e., logical). 6. Theoretical work or implemented Implemented (e.g., in the epicure system, see =-=[11]-=-). 7. Approach to addressing constraints in generation Classical constraint satisfaction. 8. Single or many constraints? How are constraints combined? Any finite number of constraints (as expressed in... |

23 | On identifying sets
- Stone
- 2000
(Show Context)
Citation Context ...(also applied in Mellish [37]). 9. Any other comments on the paper Essentially the same approach was later used for the generation of (distributive or collective) references to sets by Matthew Stone (=-=[56]-=-). Gardent’s ([16]) use of constraints in generation of referring expressions (GRE) is different because there, the constraints correspond with general GRE principles. 4.4 van Deemter [57], 2004, ‘Tow... |

22 | 2001. Towards a computational model of poetry generation - Manurung, Ritchie, et al. - 2001 |

18 | Applying explanation-based learning to control and speeding-up natural language generation
- Neumann
- 1997
(Show Context)
Citation Context ...31, 30, 29, 36, 10, 47, 59] 3.2 Units to which constraints apply Phonetic [35, 34] Punctuation markers [59] Words [24, 10, 30, 36, 44, 59] Mappings from conceptual input to lexical items [15] Phrases =-=[24, 44, 42, 43]-=- (Content of) Noun Phrases/Referring expressions [16, 12] Clauses [9] Sentences [10, 29, 39, 45, 42, 43] Syntactic Surface Structure [36, 26] Trees [6] Paragraphs [24] Discourse: Text Structures/RST T... |

17 | Soft Constraint Solving and Programming: a general framework - Bistarelli |

16 | Interleaving natural language parsing and generation through uniform processing
- Neumann
- 1998
(Show Context)
Citation Context ...rview of the different approaches to constraint-based nlg together with pointers to the relevant literature. 3.1 Nature of constraints Morphological [45] Lexical Choice [15, 24, 36, 44, 59] Syntactic =-=[6, 24, 25, 10, 30, 36, 39, 44, 45, 43]-=- Syntactic (aggregation) [9] Semantic [10, 12, 57, 16, 39, 45, 43] Semantic Ambiguity [26] Punctuation [59] Discourse [28, 36, 38, 49, 59] Affect/Opinion [24, 25] Style [24, 25, 26, 35, 34, 47, 48] 10... |

15 | De-constraining text generation
- Beale, Nirenburg, et al.
- 1998
(Show Context)
Citation Context ... constraints in generation Propose algorithm for solving finite-domain hard constraints. 8. Single or many constraints? How are constraints combined? – 9. Any other comments on the paper Beale et al. =-=[5]-=- is a precursor to work along these lines. They apply a special purpose constraint-directed control architecture called Hunter-Gatherer (HG) to sentence planning. 4.2 Callaway and Lester [9], 1997, ‘D... |

12 | A head-driven approach to incremental and parallel generation of syntactic structures
- Neumann, Finkler
- 1990
(Show Context)
Citation Context ...e aim to provide is a quick overview of the different approaches to constraint-based nlg together with pointers to the relevant literature. 3.1 Nature of constraints Morphological [45] Lexical Choice =-=[15, 24, 36, 44, 59]-=- Syntactic [6, 24, 25, 10, 30, 36, 39, 44, 45, 43] Syntactic (aggregation) [9] Semantic [10, 12, 57, 16, 39, 45, 43] Semantic Ambiguity [26] Punctuation [59] Discourse [28, 36, 38, 49, 59] Affect/Opin... |

12 |
Pipelines and size constraints
- Reiter
- 2000
(Show Context)
Citation Context ...[48], 2003, ‘Dialogue as Discourse: Controlling Global Properties of Scripted Dialogue’ . . . . . . . . . . 33 4.19 Power [49], 2000, ‘Planning text by constraint satisfaction’ . . . . 33 4.20 Reiter =-=[52]-=-, 2000, ‘Pipelines and Size Constraints’ . . . . . . . . . 34 4.21 Thiam Chye [10], 2005, ‘Multi-document Summarization by Generation’ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3... |

11 |
A framework for stylistically controlled generation
- Paiva, Evans
- 2004
(Show Context)
Citation Context ...al Language’ . . . . . . . . . . . . 29 4.16 Neumann and collaborators (1990–1998) on constraint-based grammars, reversible grammars and revision . . . . . . . . . . . . . . . 29 4.17 Paiva and Evans =-=[47]-=-, 2004, ‘A Framework for Stylistically Controlled Generation’ . . . . . . . . . . . . . . . . . . . . . . . . . . 32 4.18 Piwek and Van Deemter [48], 2003, ‘Dialogue as Discourse: Controlling Global P... |

10 | A 2000. Flexible Integrated Architecture for Generating Poetic Texts
- Ritchie, G, et al.
- 2000
(Show Context)
Citation Context .... . . . . . . . . . . 24 4.11 Langkilde and Knight [29, 31], 1998, ‘Generation that Exploits Corpus-based Statistical Knowledge’ . . . . . . . . . . . . . . . . 25 4.12 Manurung, Ritchie and Thompson =-=[35, 34]-=-, 2000, ‘Towards a Computational Model of Poetry Generation’ . . . . . . . . . . . . 25 4.13 Marciniak & Strube [36], 2005, ‘Beyond the Pipeline: Discrete Optimization in NLP’ . . . . . . . . . . . . ... |

8 |
Dialogue as discourse: Controlling global properties of scripted dialogue
- Piwek, Deemter
(Show Context)
Citation Context ...y how the distance is precisely calculated. 8. Single or many constraints? How are constraints combined? Two constraints. See above. 9. Any other comments on the paper – 32s4.18 Piwek and Van Deemter =-=[48]-=-, 2003, ‘Dialogue as Discourse: Controlling Global Properties of Scripted Dialogue’ 1. Name of the constraint(s) (if available) ‘Number of turns’ and ‘degree of emphasis’. 2. Brief informal descriptio... |

6 |
An Exploratory Application of Constraint Optimization in Mozart to
- Langkilde-Geary
- 2004
(Show Context)
Citation Context ... . . . . . . . . . . . . . . . 22 4.9 Kibble and Power [28], 2004, ‘Optimizing Referential Coherence in Text’ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 1s4.10 Langkilde-Geary =-=[30]-=-, 2004, ‘An Exploratory Application of Constraint Optimization in Mozart to Probabilistic Natural Language Processing’ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 4.11 Langkilde and... |

6 | Towards a probabilistic version of bidirectional ot syntax and semantics
- Deemter
- 2004
(Show Context)
Citation Context ...Based Approach to Explanation Generation’ . 16 4.3 Dale & Haddock [12], 1991, ‘Content Determination in the Generation of Referring Expressions’ . . . . . . . . . . . . . . . . . . 17 4.4 van Deemter =-=[57]-=-, 2004, ‘Towards a Probabilistic Version of Bidirectional OT Syntax and Semantics’ . . . . . . . . . . . . . . . . 18 4.5 Elhadad, McKeown and Robin [15], 1997, ‘Floating Constraints in Lexical Choice... |

4 | Generating with a grammar based on tree descriptions: A constraintbased approach
- Gardent, Thater
- 2001
(Show Context)
Citation Context ...he content of the NP): sets of properties to be expressed in a referring expression. 5. Nature of the constraints Semantic. 6. Theoretical work or implemented Implemented using the generator InDiGen (=-=[17]-=-) which uses the concurrent constraint programming language Oz. 7. Approach to addressing constraints in generation Classical constraint satisfaction. 20s8. Single or many constraints? How are constra... |

4 |
A survey of global optimization methods. http://www. cs.sandia.gov/opt/survey/main.html
- Gray, Hart, et al.
- 1997
(Show Context)
Citation Context ..., e.g., [2] for a list of resources. • OTSoft for Optimality Theory ([21]). • Problems involving soft constraints are a form of optimization problems. A survey of optimization methods can be found in =-=[19]-=-. A particularly relevant field is linear Programming (LP). LP addresses optimization problems in which the objective function and the constraints are linear ([1]). 3 There exist various (integer) lin... |

4 |
Discrete Optimization as an Alternative to Sequential Processing in nlg
- Marciniak, Strube
- 2005
(Show Context)
Citation Context ...owledge’ . . . . . . . . . . . . . . . . 25 4.12 Manurung, Ritchie and Thompson [35, 34], 2000, ‘Towards a Computational Model of Poetry Generation’ . . . . . . . . . . . . 25 4.13 Marciniak & Strube =-=[36]-=-, 2005, ‘Beyond the Pipeline: Discrete Optimization in NLP’ . . . . . . . . . . . . . . . . . . . . . . . . 26 4.14 Mellish, Knott, Oberlander and O’Donnell [38], 1998, ‘Experiments Using Stochastic S... |

4 | Noord. Self-monitoring with reversible grammars
- Neumann, van
- 1992
(Show Context)
Citation Context ...sive. Rather, what we aim to provide is a quick overview of the different approaches to constraint-based nlg together with pointers to the relevant literature. 3.1 Nature of constraints Morphological =-=[45]-=- Lexical Choice [15, 24, 36, 44, 59] Syntactic [6, 24, 25, 10, 30, 36, 39, 44, 45, 43] Syntactic (aggregation) [9] Semantic [10, 12, 57, 16, 39, 45, 43] Semantic Ambiguity [26] Punctuation [59] Discou... |

2 | Forest-based statistical sentence generation
- Langkilde-Geary
- 2000
(Show Context)
Citation Context ...son domains for representing word features, and on relationships between features of words, e.g., regulating the position of heads.] 9. Any other comments on the paper None. 4.11 Langkilde and Knight =-=[29, 31]-=-, 1998, ‘Generation that Exploits Corpus-based Statistical Knowledge’ 1. Name of the constraint(s) (if available) – 2. Brief informal description of the constraint(s) See below. 3. Type of constraint(... |