This paper describes the Sentence Planner (sP) in the HealthDoc project, which is concerned with the production of customized patienteducation material from a source encoded in terms of plans. The task of the sP is to transform selected, not necessarily consecutive, plans (which may vary in detail, from text plans specifying only content and discourse organization to fine-grained but incohesive, sentence plans) into completely specified specifications for the surface generator. The paper identifies the sentence planning tasks, which are highly interdependent and partially parallel, and argues, in accordance with [Nirenburg et al., 1989], that ' a blackboard architecture with several independent modules is most suitable to deal with them. The architecture is presented, and the interaction of the sentence planning modules within this architecture is shown. The first implementation of the sP is discussed; examples illustrate the planning process in action. 1 S e n t e n c e P l a n n i n g 1.1 I n t r o d u c t i o n Most current models of text generation include a phase of content selection and organization, usually performed by a text planner or schema application engine, followed by a phase of grammatical surface-form rendering, performed by a sentence generator. In practice, it is usually found that the sentence generator requires more detailed linguistic information than text planners or schema appliers can provide [Meteer, 1991; Rambow and Korelsky, 1992; Hovy, 1992; Panaget, 1994; Wanner, 1994]. So further planning is required. Following [Rambow and Korelsky, 1992], we call this additional planning task sentence planning (even though some operations may cross sentence boundaries). A sentence planner (sP) must specify one of the various possible alternative phrasings at roughly the sentence and clause level. By transforming and augmenting its input, the sentence planner produces representations detailed enough for the surface generator to operate deterministically. Consider an example where the lack of sentence planning results in an awkward text: (o) In some instances, an implant wears out, loosens, or fails. I f an implant wears out, loosens, or fails, it will have to be removed. More appropriate alternatives can be generated when different sentence planning techniques are used: (1) A l t e r n a t i v e r e fe rence : In some instances, an implant wears out, loosens, or fails. I f this happens, it will have to be
[1]
I. Mel.
Meaning-Text Models: A Recent Trend in Soviet Linguistics
,
1981
.
[2]
Franck Panaget,et al.
Using a textual representational level component in the context of discourse or dialogue generation
,
1994
.
[3]
James Pustejovsky,et al.
The Generative Lexicon
,
1995,
CL.
[4]
Eduard H. Hovy,et al.
On Lexical Aggregation and Ordering
,
1996,
INLG.
[5]
Leo Wanner.
Building Another Bridge over the Generation Gap
,
1994,
INLG.
[6]
Sergei Nirenburg,et al.
Controlling a Language Generation Planner
,
1989,
IJCAI.
[7]
David D. McDonald.
Subsequent reference: syntactic and rhetorical constraints
,
1978,
TINLAP '78.
[8]
Christian M. I. M. Matthiessen,et al.
Text Generation and Systemic-Functional Linguistics: Experiences from English and Japanese
,
1992
.
[9]
Manfred Stede,et al.
Ma(r)king concessions in English and German
,
1995,
ArXiv.
[10]
Agnès Tutin,et al.
Lexical choice in context: generating procedural texts
,
1992,
COLING.
[11]
Manfred Stede.
A generative perspecl: ive on verbs and their readings
,
1996,
INLG.
[12]
Michael Zock,et al.
Trends in Natural Language Generation An Artificial Intelligence Perspective
,
1996,
Lecture Notes in Computer Science.
[13]
Graeme Hirst,et al.
A Computational Theory of Goal-Directed Style in Syntax
,
1993,
Comput. Linguistics.
[14]
James H. Martin,et al.
Expressing Rhetorical Relations in Instructional Text: A Case Study of the Purposes Relation
,
1995,
Comput. Linguistics.
[15]
Eduard H. Hovy,et al.
Aggregation in Natural Language Generation
,
1993,
EWNLG.
[16]
Roland Kasper.
SPL: A sentence plan language for text generation
,
1989
.
[17]
M. Meteer.
Bridging the generation gap between text planning and linguistic realization
,
1991
.
[18]
Douglas E. Appelt.
Planning Natural-Language Utterances
,
1982,
AAAI.
[19]
Leo Wanner,et al.
A collocational based approach to salience-sensitive lexical selection
,
1990,
INLG.
[20]
Owen Rambow,et al.
Applied Text Generation
,
1992,
ANLP.
[21]
Robert Dale,et al.
Computational Interpretations of the Gricean Maxims in the Generation of Referring Expressions
,
1995,
Cogn. Sci..
[22]
John A. Bateman.
Upper Modeling: organizing knowledge for natural language processing
,
1990,
INLG.
[23]
Helmut Horacek,et al.
An Integrated View of Text Planning
,
1992,
NLG.