Generating anaphoric expressions [Elektronische Ressource] : contextual reasoning in sentence planning / vorgelegt von Kristina Striegnitz

Sujets

• Informatik

Informations

• Publié par: universitat_des_saarlandes
• Publié le: 01 janvier 2005
• Nombre de lectures: 2
• Langue: English

Extrait

Generating Anaphoric
Expressions
Contextual Reasoning in Sentence Planning
Dissertation
zur Erlangung des akademischen Grades eines Doktors der Philosophie
der Philosophischen Fakult˜ aten der Universit˜ at des Saarlandes
vorgelegt von
Kristina Striegnitz
aus Berlin
Saarbruc˜ ken, 2005
Diese Dissertation wurde im Rahmen eines kooperativen Promotionsverfahrens
(cotutelle de th?ese) zwischen der Universit˜ at des Saarlandes und der Univer-
sit˜ at Henri Poincar¶e (Nancy, Frankreich) erstellt.Dekanin: Prof. Dr. Wolfgang Schweickard
Berichterstatter: Prof. Dr. Manfred Pinkal, Universit˜ at des Saarlandes
Dr. Claire Gardent, CNRS, Nancy
Tag der letzten Prufungsleistung:˜ 19.11.2004Abstract
This thesis investigates the contextual reasoning involved in the production of
anaphoric expressions in natural language generation systems. More specifl-
cally, I propose generation strategies for two types of discourse anaphora which
have not been treated in generation before: bridging descriptions and additive
particles. To this end the contextual conditions that govern the use of these
expressions have to be formalized. The formalization that I propose is based
on notions from linguistics and extends previous approaches to the generation
of co-referential anaphora. I then specify the reasoning tasks that have to be
carried out in order to check the contextual conditions. I describe how they
can be implemented using a state-of-the-art reasoning system for description
logics, and I compare my proposal to alternative approaches using other kinds
of reasoning tools. Finally, I describe an experimental implementation of the
proposed approach.
R¶esum¶e
Cette th?ese porte sur le raisonnement contextuel impliqu¶e par la produc-
tion d’expressions anaphoriques dans un syst?eme de g¶en¶eration de langue na-
turelle. Plus pr¶ecis¶ement, nous proposons des strat¶egies pour g¶en¶erer deux
types d’anaphore n’ayant encore jamais ¶et¶e consid¶er¶es dans la g¶en¶eration :
les anaphores associatives et les adverbes additifs. Dans ce but, nous for-
malisons tout d’abord les conditions contextuelles d¶eterminant l’usage de ces
expressions. Cette formalisation est bas¶ee sur des notions linguistiques, et
¶etend les approches ant¶erieures de g¶en¶eration d’anaphores co-r¶ef¶erentielles.
Ensuite, nous sp¶eciflons les t^aches de raisonnement a? efiectuer pour v¶erifler
ces conditions contextuelles. Nous d¶ecrivons comment ces t^aches peuvent ^etre
impl¶ement¶ees en utilisant un syst?eme d’inf¶erence pour les logiques de descrip-
tion, et nous comparons notre approche a? des approches alternatives utilisant
d’autres syst?emes d’inf¶erence. Pour flnir, nous d¶ecrivons une impl¶ementation
exp¶erimentale de notre approche.
iAcknowledgments
First of all, I would like to thank my supervisors Claire Gardent and Manfred
Pinkal without whom this thesis would not have been possible. I am grateful
to Claire for all the things that I learned from her, for always trusting me to
do my work properly, and for somehow getting me to flnally write my thesis
in the end.
I am very grateful to Jean-Marie Pierrel for sorting out all kinds of admin-
istrative problems. The French and the German regulations on how to write
and defend a thesis are inconsistent | that I nevertheless managed to satisfy
both is largely due to Jean-Marie.
I also thank Bonnie Webber for making it possible for me to spend some
time in Edinburgh and for getting me started on the additive particles.
A very special thanks go to Carlos Areces for bribing me into writing, for
reading and commenting on various versions of my thesis, for making sure that
in the end all my references have page numbers, and for loads of other things.
H¶el?ene Manu¶elian and Garance Paris read the French parts of my thesis
and turned them into something closely related to real French.
Finally, I thank everybody who has in some way or another contributed to
making the past four years a very happy part of my life.
iiiContents
Die Generierung anaphorischer Ausdruc˜ ke (Zusammenfassung) 1
R¶esum¶e d¶etaill¶e en fran» cais 9
1 Introduction 19
Part I Background 27
2 Natural Language Generation 29
2.1 NLG Subtasks . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.1.1 Document Planning . . . . . . . . . . . . . . . . . . . . 30
2.1.2 Microplanning . . . . . . . . . . . . . . . . . . . . . . . 34
2.1.3 Linguistic Realization . . . . . . . . . . . . . . . . . . . 35
2.2 Architectures . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
2.3 The Spud System . . . . . . . . . . . . . . . . . . . . . . . . . 37
2.3.1 Discourse Context . . . . . . . . . . . . . . . . . . . . 38
2.3.2 Communicative Goal . . . . . . . . . . . . . . . . . . . 39
2.3.3 Grammar . . . . . . . . . . . . . . . . . . . . . . . . . 39
2.3.4 Search Problem . . . . . . . . . . . . . . . . . . . . . . 41
2.3.5 Textual Economy . . . . . . . . . . . . . . . . . . . . . 43
3 Generating Anaphora 47
3.1 Generating Deflnite Noun Phrases . . . . . . . . . . . . . . . . 47
3.1.1 Dale and Reiter’s Algorithm for the Generation of Defl-
nite Descriptions . . . . . . . . . . . . . . . . . . . . . 48
3.1.2 Deflnite in Spud . . . . . . . . . . . . . . 53
v3.1.3 Generating Pronouns . . . . . . . . . . . . . . . . . . . 54
3.2 Other Kinds of Anaphora . . . . . . . . . . . . . . . . . . . . 56
3.2.1 Verb Phrase Ellipsis . . . . . . . . . . . . . . . . . . . 56
3.2.2 Non-sentential Answers in Dialogue . . . . . . . . . . . 57
3.2.3 Non-canonical Syntactic Constructions . . . . . . . . . 58
Part II Generating Anaphoric Expressions 61
4 Beyond Co-reference: Bridging Descriptions 63
4.1 Bridging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
4.1.1 Bridging Descriptions . . . . . . . . . . . . . . . . . . . 64
4.1.2 Anchors . . . . . . . . . . . . . . . . . . . . . . . . . . 66
4.1.3 Bridging Relations . . . . . . . . . . . . . . . . . . . . 67
4.2 Familiarity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
4.3 Uniqueness . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
4.4 Generating Bridging Descriptions . . . . . . . . . . . . . . . . 76
4.4.1 An Incremental Algorithm for Generating Bridging De-
scriptions . . . . . . . . . . . . . . . . . . . . . . . . . 76
4.4.2 AdaptingSpud to the Generation of Bridging Descriptions 81
4.5 Probable and Inducible Parts . . . . . . . . . . . . . . . . . . 89
4.6 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
4.6.1 Summary . . . . . . . . . . . . . . . . . . . . . . . . . 92
4.6.2 Further Questions . . . . . . . . . . . . . . . . . . . . . 93
5 Beyond NP Anaphora: Additive Particles 95
5.1 Approaches to Additive Particles . . . . . . . . . . . . . . . . 96
5.1.1 Additive Particles as Presupposition Triggers or Context
Markers . . . . . . . . . . . . . . . . . . . . . . . . . . 96
5.1.2 Alternative Particles as Focus Particles . . . . . . . . . 98
5.2 Deflning Also-Parallelism . . . . . . . . . . . . . . . . . . . . . 103
5.3 Alternative Sets in the Literature . . . . . . . . . . . . . . . . 106
5.3.1 Ontological Alternatives . . . . . . . . . . . . . . . . . 106
5.3.2 Discourse Structure based Alternatives . . . . . . . . . 108
5.4 Difierent Sources for Alternatives . . . . . . . . . . . . . . . . 112
5.5 Deriving Alternative Sets in a Generation System . . . . . . . 117
5.5.1 Discourse Structure Based Alternative Sets . . . . . . . 117
5.5.2 Explicitly Introduced Groups and Ontology Based Al-
ternative Sets . . . . . . . . . . . . . . . . . . . . . . . 119
5.6 Generating Additive Particles . . . . . . . . . . . . . . . . . . 119
5.6.1 Spud with Marking Principles . . . . . . . . . . . . . . 119
5.6.2 The Additive Principle and Additive Markers 121
5.6.3 An Example . . . . . . . . . . . . . . . . . . . . . . . . 123
vi5.7 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
5.7.1 Summary . . . . . . . . . . . . . . . . . . . . . . . . . 124
5.7.2 Further Questions . . . . . . . . . . . . . . . . . . . . . 125
Part III Implementation 131
6 Automated Theorem Proving for Contextual Reasoning 133
6.1 The Reasoning Tasks . . . . . . . . . . . . . . . . . . . . . . . 133
6.2 Using a Description Logic Reasoner . . . . . . . . . . . . . . . 135
6.2.1 Computing Speaker and Hearer Anchors . . . . . . . . 136
6.2.2 Checking the Familiarity and Uniqueness Conditions . 139
6.2.3 Computing Alternative Sets . . . . . . . . . . . . . . . 140
6.2.4 Checking for Also-parallel Eventualities . . . . . . . . . 141
6.3 First Order Logic Theorem Provers . . . . . . . . . . . . . . . 141
6.4 Model Checking . . . . . . . . . . . . . . . . . . . . . . . . . . 144
6.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
7 An Example Application 147
7.1 A Text-based Computer Game . . . . . . . . . . . . . . . . . . 147
7.1.1 Overview of the Game Architecture . . . . . . . . . . . 148
7.1.2 The World Model . . . . . . . . . . . . . . . . . . . . . 149
7.1.3 The Discourse Model . . . . . . . . . . . . . . . . . . . 151
7.1.4 The Language Analysis Module . . . . . . .