DISCOURSE MARKER CHOICE 
IN SENTENCE PLANNING 
Brigitte Grote* 
Otto-von-Guericke Universit~it Magdeburg 
Institut fiir Wissens- und Sprachverarbeitung 
-Abstract 
Manfred Stede t 
Technische Universit~t Berlin 
Projektgruppe KIT 
In text, discourse markers signal the kind of coherence relation holding between adjacent text 
spans; for example, because, since; and for this reason are different markers for causal relations. 
For any but the most simple applications of text generation, marker Selection is an important 
aspect of producing cohesive text. However, present systems use markers in fairly simplistic ways 
andcannot make use of the full potential of markers that language offers for a given relation. 
• To improve this situation, we propose a specialized lexicon for discourse markers, which holds 
the relevant constraints and preferences associated with the markers, and which can be used by 
a text generator to make an informed choice among alternative ways of expressing a relation 
in the given context, we demonstrate how .the lexicon can be employed in the generation 
• process and propose to perform discourse marker choice in the sentence planning stage, where 
the interactions with other generation• decisions can be accounted for. 
1 Introduction 
When a coherence relation ties two adjacent portions of text together, it is often lexically signalled 
on the linguistic surface with a suitable word--most Often a conjunction, but also a preposition, a 
prepositional phrase or an adverb \[Quirk et al. 1972\]. The set of words from these grammatically 
heterogeneous classes that can signal coherence relations we call discourse markers. For example, 
:the CONCESSION relation in English can be signalled with the subordinator although, the adverb 
Still, the preposition despite, and a number of other words. 
For most coherence relations, language offers quite a variety of such markers, as several studies 
Of individual relations have demonstrated (see references in Section 2). Accordingly, from the 
generation perspective, a serious choice task arises if the produced • text is not only to simply signal 
:the coherence relation, but moreover to reflect pragmatic goals, stylistic considerations, and the 
different connotations markers have. The importance of these factors was stressed by Scott and de 
Souza \[1990\], who gave a number of informal heuristics for when and how to signal the presence of 
coherence relations in text. Fleshing out the choice task in order to come up with a computational 
model, though, reveals two sources of difficulty. 
For one thing, in addition to syntactic variety, the precise semantic and pragmatic differences 
between similar markers can be quite difficult to determine. For instance, the CONCESSION markers 
although and even though differ merely in emphasis; the CAUSE markers because and since differ 
in whether they mark the following information as given or not; the German CAUSE markers weil 
and denn differ in the illocution type of the conjuncts (proposition versus statement). Second, 
,Complete address: Otto-von-Guericke Universit~it Magdeburg, IWS/FIN, Universit~tsplatz 2, 39106 Magdeburg, 
Germany; email: grote@iws, cs.tmi-magdeburg, de " 
: tComplete address: Technische Universit~t Berlin, Fachbereich Inforraatik, Projektgruppe KIT, Sekretariat FR 
6-10, Franklinstr. 28/29, 10587 Berlin, Germany; email: stedeecs.tu-berlin.de 
128 
I 
I 
I 
I 
I 
I 
.Jl 
! 
ii 
I 
.I 
I 
.I 
I 
I 
I 
I 
I 
I 
I 
I 
I 
I 
I 
I 
I 
I 
I 
I 
I 
i 
I 
I 
I 
the dependencies between marker choice and other generation decisions are rather intricate. The 
idea of avoiding them is, presumably, the reason for the simplistic treatment of marker choice in 
typical generators to-date: They regard discourse markers as mere consequences of the structural 
decisions, hence do not perform any choice. We wish to demonstrate, however , that this strategy, 
which is typical for dealing with closed-class lexical items in general, is too great .a simplification 
in these cases. 
In this paper, we propose to use a discourse marker lexicon as a declarative resource for the 
sentence planning stage of the text generation process. The paper is organized as follows. Section 
2 examines the role of discourse markers in NLG and reviews the state of the art. Section 3 briefly 
summarizes the ideas on sentence planning that have arisen in the past few years and argues that 
for a sophisticated treatment of discourse marker choice, a dedicated lexicon is to be used as one 
information resource in sentence planning. Section 4 •introduces the discourse marker lexicon we 
are currently developing, and Section 5 describes how this lexicon can be usefully employed in the 
sentence planning phase to realize more flexible marker production. 
.J _ 
2 Discourse markers in NLG 
We follow Moser and Moore \[1995\] in assuming that three distinct though interrelated decisions 
have to be made when generating discourse markers: Whether to place a marker or not (marker 
occurrence), where to place a marker (marker placement), and finally, which marker to use (marker selection). 
Research on connectives in the context of NLG has focused on the selection of markers to 
produce coherent and cohesive multi-sentential text. Studies fall into two distinct groups: First, 
studies are concerned with identifying the characteristic •properties of a small set of similar markers, 
and determining the reasons behind choosing a particular marker from this set in a given context; 
examples are the markers since and because \[Elhadad and McKeown 1990\], or the temporal mark, 
ers before • and while \[Dorr and Gaasterland 1995\]. Second, a number of studies take particular 
(RST-)relati0ns as a starting-point, and examine how these relations are signalled on the linguistic 
surface; examples are the PURPOSE, RESULT and PRECONDITION relations \[Vander Linden 1994\], 
the CONCESSION relation \[Grote et al. 1997\], and the subject-matter relations occurring in a tech- 
nical domain \[RSsner and Stede 1992, Delin et al. 1996\]. However , these are all isolated studies, 
geared towards a particular application. There is at present no overall framework that supports 
informed and motivated marker generation for more than a small set of markers and relations. 
The broadest overview on discourse markers to our knowledge is the descriptive work of Knott 
and Mellish \[1996\], but it does not specifically address the NLG perspective. Moser and Moore 
\[1995\] and DiEugenio et al. \[1997\] also take a broader view on marker production in that they try 
to determine general factors that influence the use of markers in text, and in that they consider 
more than pairs of propositions.• However, they are largely • concerned with marker occurrence and 
placement, not with marker selection. 
3 Sentence planning 
The traditional §plit of NLG systems in a content determination/what-to-say component and a 
realization/how-to-say component was in recent years •supplemented by an intermediate stage: 
sentence planning, sometimes called micro-planning (e.g., Rambow and Korelsky \[1992\]). The 
primary motivation for this step is to relieve the text planner from language-specific knowledge, • 
and to relieve the realization component from any planning or decision-making that potentially 
• 129. 
affects the mean.ing of the utterance. Hence, better control of the overall generation process is 
gained. We do not elaborate the advantages further here; see, for example, \[Panaget 1994\]. 
What are the specific decisions to be made by the sentence planner? We think it is important to 
separate the format{re decisions from the motivations that lead to the particular choices. Following 
Wanner and Hovy \[1996\], a sentence planner has tomake the following decisions: Fine-grained 
discourse structuring, including discourse marker choice; sentence grouping and sentence content 
determination; clause-internal structuring; Choice of referring expressions; lexical choice} Two 
groups of considerations are important for these tasks: First, the motivating factors such as stylistic 
choices, semantic relations, intentions, theme development, focusing, discourse history. Second, 
the interactions with other desicions, because different formative decision may realize the same 
motivation. In contrast to present NLG systems, which realize the production of marker choices as 
a mere consequence of other sentence level decisions, we think that sentence planning interactions 
ough t to be respected for discourse markers, too, as the following examples illustrate: 
• Ordering of related clauses (cause-effect vs. effect-cause) " . 
Because he was unhappy, he asked to be transferred, vs. He asked to be transferred, for he 
was unhappy, vs. • For he was unhappy, he asked to be transferred. 
• Aggregation 
He has quarrelled with the chairman. He resigned from his post. vs. He has quarrelled with 
the chairman and resigned from his post. 
• Delimit individual sentences 
They fought a battle. Afterwards, it snowed, vs. After they fought a battle, it snowed. 
• • Clause-internal structuring (hypotaxis vs. parataxis) 
Although he has tried hard, he failed, vs. He tried hard, but he failed. 
• Lexical choice (to know vs. ignorance) 
She died, because she didn't know the rules, vs. She died through ignorance of the rules. 
• Realizing negation 
He will not attend unless he finishes his paper, vs. He will attend if he finishes his paper. 
Due to these interdependencies, any fixed order of making decisions in sentence planning wiil 
impose limitations on the expressiveness of the system. Accordingly, we advocate a flexible order 
of decisiommaking, as it can be realized in a blackboard-based architecture such as proposed 
by DIOGENES \[Nirenburg et al. 1989\] and HealthDoc \[Wanner and Hovy 1996\]. Moreover, the 
individual modules or knowledge sources should rely on declarative representations as much as 
possible; otherwise the control process becomes extremely complicated. And one of the declarative 
sources of information, we feel, should be a lexicon that assembles •specifically the information 
associated with discourse markers. 
4 ••The Discourse marker lexicon 
4.1 Discourse markers as lexical entities 
The traditional clistinction between content words and function words (or open-Class and closed- 
class items) relies on the stipulation that the former have their "own" meaning independent of the 
1How exactly these tasks axe to be accomplished depends on the nature of the input and output representations, 
and thus on the architecture of the generator. In Section 5, we will introduce the framework we axe using and 
characterize the integratio n of marker choice into the sentence planning phase. 
•130 
I 
I 
I 
I 
,I 
! 
i 
{I 
ii 
,| 
!. 
i 
context in which they are used, whereas the latter assume meaning only in context. Then, content 
words are assigned to the realm of the lexicon, whereas function words are treated as a part of 
grammar. For dealing with discourse markers, we do not regard this distinction as particularly 
helpful, though. These words can carry a wide variety of semantic and pragmatic overtones, which 
render the task of selecting a marker meaning-driven, as opposed to a mere consequence of structural 
decisions. 
Furthermore, notice that a number of lexical relations customarily used to assign structure to 
the universe of "open Class" lexical items can be applied to discourse markers as well. For instance, 
the German words obzwar and obschon (both more formal variants of obwohl = although) are at 
least very close to being synonyms. As for plesionyms (near-synonyms), although and though, 
according to Martin \[1992\], differ in formality, and although and even though differ in terms of 
emphasis. If and unless can be seen as antonyms, as they both express conditionality, but with 
opposite polarity. Some markers are more specific than others, thus display hyponymy. E.g., 
but can signal a general CONTRAST or a more specific CONCESSION. Finally, other than being 
more or less specific, some markers can signal quite different relations; e.g., while can be used for 
TEMPORAL CO-OCCURRENCE, and also for CONTRAST. Hence, the marker is polysemous. 
For these reasons, discourse markers should be described by a dedicated lexicon that provides a 
classification of their syntactic, semantic and pragmatic features and characterizes the relationships 
between similar markers. This will be a lexicon whose main grouping criterion is function rather 
than grammatical category; not surprisingly, this is motivated by the production perspective, where 
the parameters governing the generation decisions play the central role. 
4.2 Methodology 
Methodological considerations pertain to the two tasks of determining the set of words we regard 
as discourse markers, and determining the lexical entries for these words. 
Finding the "right" set of discourse markers is not an easy task, since the common lexicographic 
practice of having syntactic behaviour as the criterion for inclusion does not apply. Knott and 
Mellish \[1996\] provide an apt summary of the situation. Their 'test for relational phrases' is a 
good start, but geared towards the English language (we are investigating German as well), and 
furthermore it catches only items relating clauses; in Despite the heavy rain, we went for a walk it 
would not detect a cue phrase. To identify more markers, we worked with traditional dictionaries 
and with grammars like Quirk et al. \[1972\] and Helbig and Buscha \[1991\]. The resulting set of 
markers is further validated by investigating coherence relations and their possible realizations; 
here, we can draw on our earlier work \[RSsner and Stede 1992, Grote et al. 1997\]. 
As for the shape of the lexical entries, there are two tasks: First, determining the distinguishing 
features and classifying markers according to these features, and second, finding appropriate com- 
putational representations. At present, we axe mostly concerned with the first step, but in section 
5, we make an initial proposal for representations. 
Regarding the set of features, our goal can be characterized as finding a synthesis of two different 
perspectives on marker description, between which there has been little overlap in the research 
literature: Text linguistics considers markers as a means to signal coherence, and provides us with 
insights on the semantic and pragmatic properties of marker Classes, hence approaches the matter 
"top-down". On the other hand, grammars and style guides provide syntactic, semantic and stylistic 
properties of individual markers, thus look "bottom-up". 
Specifying the distinctions within sets of similar markers can be quite subtle. In addition to 
drawing on our earlier work cited above, we employ techniques such as paraphrasing, Knott's 
substitution test \[Knott and Mellish 1996\], analysis of typical distributions using corpora, and con- 
trastive studies. Extracting features in this way seems justified since at this stage we arc unlike 
131 
feature unless for however even though notwithstanding 
syntactic 
part-of-speech 
connection type 
scope of marker 
position 
linear ordering 
semantic 
semantic relation 
polarity 
functional order 
nuclearity 
pragmatic 
formality 
emphasis 
discourse relation 
subordinator 
hypotaxis 
S simple 
front 
N S 
condition 
act negated 
any 
N: act 
standard 
none 
PRECONDITION 
subordinator 
hypotaxis 
any 
front 
NS 
cause 
any 
effect-cause 
N: effect 
standard 
none 
VOL.CAUSE 
adverb 
intersent. 
any 
front/mid/end 
SN 
concession 
any 
conceding-conceded 
N- conceded 
standard 
none 
CONCESSION 
subordinator 
hypotaxis 
S simple 
front 
any 
concession 
any 
any 
N: conceded 
standard 
intensified 
CONCESSION 
preposition 
intraclausal 
N and S simple 
front 
any 
concession 
any 
any 
N: conceded 
formal 
none 
CONCESSION 
Table 1: Sample lexicon entries for some English markers 
DiEugenio et al. \[1997J--not concerned with the predictive power of individual features but rather 
with decomposing markers into features that are relevant for integrating marker choice into sentence 
planning. 
4.3 The shape of the lexicon 
The initial set of features we have thus obtained can be grouped in the traditional categories: 
Syntactic features are the part-of:speech of a marker and the type of connection it establishes 
(prepositions link constituents within a clause; conjunctions build a paratactic or hypotactic struc- 
ture, but some can also function as intersentential linkers). The scope of a marker is the complexity 
of the segments it can combine (complex subtree or simple propositions). The linear ordering of 
the conjuncts can differ from marker to marker (e.g., with the connective for, the subordinate 
clause is always postponed ) as well as the marker's position within the segment (e.g., prepositions 
always occur at the beginning of a segment; adverbs like however can occur in front-, mid- and 
end-position). 
Semantic features are foremost the semantic relation established (e.g. causal or temporal 
link). Some markers show a particular behaviour towards negation, which is related to polarity 
(e.g., ffversus unless). Further, we observe that Certain markers impose what we term a functional 
ordering, for instance, for requires the order effect-cause. 
Pragmatic features include the discourse relation expressed by the marker and the type of 
illocutionary acts it conjoins (e:g., German well links propositions, denn links judgements). Some 
markers differ in terms of presuppositions and the assignment of given/new (e.g., because versus 
since). Stylistic features represent dimensions like formality and emphasis. 
To illustrate how these features discriminate between markers, table 1 gives five preliminary 
sample entries. N is a shorthand for nucleus in the RST sense, S for satellite. Notice that table 
1 is merely an illustration, and not all the features introduced above are actually required for 
classification. Issues of lexical representation, including dealing with polysemy and homonymy in 
some inheritance-based formalism, will be addressed in a later stage of the project. 
132 
~N'D. 
I~~ II01 II11. i141 1151 
131 141 171 St CE 
181 191 
Figure 1: Input structure to sentence planner 
5 The discourse marker lexicon in sentence planning 
Having outlined the discourse marker lexicon as a general resource, we now turn to the question of 
using it in sentence planning. Even though the lexicon is still under development, we will illustrate 
with several prototypical representations how a sentence planner can exploit the various realization 
options offered by the lexicon. 
We assume the following framework: a discourse structure tree loosely based on RST \[Mann 
and Thompson 1988\] serves as input to the sentence planner. RS-trees comprise a set of proposi- 
tions as leaf nodes; the internal nodes represent coherence relations holding between the daughter 
nodes. The tree is encoded in the description logic LOOM \[MacGregor and Bates 1987\], and the 
propositions are represented following the ontology used in the MOOSE system \[Stede 1996\]. The 
nature Of these representations need not concern us here, but it is important thatthey are all 
"grounded" in a knowledge base so that type checking via subsumption can take place. 
The output of the sentence planning module is a sequence of lexicalised sentence-semantic 
specifications (SemSpecs), based on SPL \[Kasper 1989\]. Accordingly, sentence planning in this 
framework amounts to linearizing a discourse representation tree. As front-end sentence generator, 
we use KPML \[Bateman 1997\]. A sample input structure from the domain of maintenance manuals 
is given in figure I; figure 2 shows one possible realization. Numbers in the tree correspond to text 
segments, and each segment corresponds to one underlying proposition. 
\[Wait\]l until \[the engine is cool\]2, then \[turn the radiator cap clockwise\]3 until \[it stops\]4. \[DO 
NOT PRESS DOWN WHILE TURNING THE CAP\]5. After \[any remaining pressure has 
been relieved\]6, \[remove the cap\]7 by \[pressing down\]8 and \[again turning it counterclockwise\]9. 
\[Add enough coolant\]10 to \[fill the radiator\]ll, and \[reinstall the cap\]12. \[Be sure to tighten it 
securely\]13. \[Fill the reserve tank up to the max mark\]14 with \[the engine cold\]15. 
• Figure 2: One linguistic realization of the RST-tree. 
5.1 The "generation view" of the discourse marker lexicon 
From the production perspective, the lexical features are to be classified with respect to when and 
where they come into play in the generation process; this amounts to one particular "view" on the 
information coded in the lexicon. We propose these categories: 
• Applicability conditions: The necessary conditions that need to be present in the input 
representation for the marker to be a candidate. Chiefly, this is the semantic/discourse 
relation to be expressed, and also (if applicable) features pertaining to presuppositions and 
intentions. 
133 
• Combinability conditions: The constraints that the marker imposes on its neighbouring 
linguistic constituents (the 'syntagmatic' dimension). These are syntactic constraints on sub- 
categorization and semantic type constraints, which interact with other realization decisions 
in sentence planning. 
• Distinguishing features: If preferential choice dimensions, such as style, brevity, etc., are 
attended to in the •system, then these features serve to distinguish markers that are otherwise 
(nearly) synonymous (the 'paradigmatic' dimension). 
For encoding this information, we adopt the framework used in the lexicalization approach of 
the MOOSE sentence generator \[Stede 1996\]. Here, lexicon entries consist of (inter alia) the three 
zones denotation, partial SemSpec (PSemSpec), and stylistic features. The •denotation is the part to 
be matched (qua subsumption) against the input rePresentation; it may contain type restrictions. 
The PSemSpec is an SPL-like template that includes a : lex annotation with the actual lexeme and 
possibly variables that are replaced by other PSemSpecs in the course of the lexicalization process. 
Also, any realization directives needed by the front-end generator are stated here. Stylistic features 
are used for preferential choice between words that would all be applicable in a particular context. 
When generalizing this framework to include discourse markers (and hence allowing for produc- 
ing complex sentences), the denotation of a marker would be an RST relati0n 2 with variables for 
the relata, possibly enriched with type constraints. For relations with a nuCleus and a satellite, we 
always write them in this order, hence (RELATION NUCLEUS SATELLITE). As a simple case, con- 
sider the subordinating conjunction until, which we take to be a marker of the relation UNTIL 3, a 
straightforward case indeed. Its denotation is (UNTIL X (STATE Y)), meaning that it can be used 
to verbalize any UNTIL node whose satellite is of type STATE, according to the ontology or domain 
model in the knowledge base. 
The variables used in the denotation also :appear in the PSemSpec of until, so that partial 
SemSpecs canbe combined together correctly. Here, the nucleus of the UNTIL relation becomes 
the domain of the rst-until relation as defined in the KPML Upper Model, 4 and the satellite 
is mapped to range, which we further constrain to be a relational-process (in Upper Model 
terms). Furthermore, we add :theme X to ensure that the nucleus is ordered before the satellite 
(to avoid until Y, X). The complete lexicon entry together with a few more exa=mples is given in 
figure 3: The denotations and PSemSpecs for the subordinating conjunctions until marking UNTIL 
and after, if, then, unless marking PRECONDITION, and for the preposition with in its function as 
marker for the relation PRECONDITION. 
5.2 Marker choice 
In MOOSE, lexical options constitute the search space for building a lexicalized semantic sentence 
specification. Now, we generalize this idea to discourse trees: For propositional nodes MOOSE 
calculates all possible lexical options; for coherence relation nodes, the list of options realizing the 
node is taken from the discourse marker lexicon by matching the node against the applicability 
conditions of the lexicon entries. Thus, the entire discourse tree is annotated with verbalization 
options, which together constitute the search space for sentence planning. 
~The relations used in denotations effectively constitute the interface between the lexicon and the text planner 
producing the discofirse tree. At present we use RST, but we regard this only as an interim solution. For the purposes 
of this paper, the precise inventory of relations used is not critical. 
3The discourse relation UNTIL was introduced by RSsner and Stede \[1992\]; its status is somewhat questionable, 
but since we do not address the general issue of discourse relations here, we simply assume its existence. 
4All Upper Model discourse relations bear the prefix 'RST', which at this point unfortunately might produce 
confusion about the variety of relations under discussion. 
134 
I 
I I/ 
I 
I 
UNLESS 
:DEN (PRECONDITION X Y) 
:PSEM (r / rst-precondition 
:domain (p / (process X) 
:polarity negative) 
:range Y 
:theme X) 
UNTIL 
:DEN (UNTIL X (state Y)) 
:PSEM (r / rst-until 
:domain X 
:range (Y / relational-process) 
:theme X 
:lex until) 
THEN 
:DEN (PRECONDITION X ¥) 
:PSEM (X / proces s ) 
(Y / process 
:conjunctive precondition 
:lex then) 
WITH 
:DEN (PRECONDITION X (state Y)) 
:PSEM (p / (process X) 
:precondition (r / (property-ascription Y) 
:lex with)) 
IF 
:DEN (PRECONDITION XY) 
:PSEM (r / rst-precondition 
:domain X 
" :range (p /. (process Y) 
:tense present) 
:lex if) 
AFTER 
:DEN (PRECONDITION X (activity Y)) 
:PSEM (p / posterior 
:domain X 
:range (p / (process Y) 
:tense .\[perfect, past perfect\]) 
:lex after) 
Figure 3: Sample (partial) lexicon entries 
To illustrate this approach, consider the propositions 14 and 15 in the sample text: Fill the 
reserve tank with the engine cold. Here, the PRECONDITION relation is Signalled by the in- 
traclausal linker with (see the lexicon entry above). Other realizations of this RS subtree are 
(\[Vander Linden 1994\] offers a similar range): 
1. If the engine is cold, fill the reserve tank up to the max mark. 
2. When the engine is co/d, fill the reserve tank up to the max mark. 
3. Fill the reserve tank up to the max mark, only if the engine is cold. 
4. After the engine has cooled down, fill the reserve tank to the upper mark. 
5. Do not fill the reserve tank (up to the max mark) unless the engine is cold. 
6. Make sure that the engine is cold. Then, fili the reserve tank up to the max mark. 
To arrive at variant formulations of this kind, depending on different parameters and/or context, 
our first step is to set up the search space of verbalization options. While MOOSE performs this 
step for propositions, we will here focus on the coherence relation nodes. In our example, the 
marker lexicon yields a set of markers that match the applicability condition (PRECONDITION X 
Y): after, if, only if, then, unless, when and with. These are annotated at the node, as shown in 
figure 4, where the leaf nodes are annotated with (shorthands for) some of the lexical options found by 
MOOSE. 
.~, RST-PRECONDITION / ~ f'Only if'then'unless'when'with\] 
1141 |lS\] nucleus 
satellite \[activity:fill\]\] \[activity: cool down\] 
\[state: cold\] 
Figure 4: Annotated subtree for \[14\] \[15\] 
135 
From this search space, different decisions made by sentence planning "expert" modules lead 
to different verbalizations. For instance, assume that the sentence-structuring expert calls for a 
hypotactic structure; this is satisfied by PSemSpecs of the form: (r / rst-precondition :domain 
X : range Y), hence by the markers if, only if, unless and when. If the clause-ordering expert calls 
for the order satellite-nucleus, unless is ruled out as it requires the nucleus to be stated first (see 
the lexicon entry below). The remaining choice between only if, ffand when is left to fine-grained 
discrimination (e.g., only if is more emphatic), which we do not elaborate here. 
Alternatively, assume that the sentence-delimitation expert posits that the relation be expressed 
• in two separate sentences. As a consequence, the ordering is satellite-nucleus. These constraints 
- are satisfied only by the marker then (example 6). The sequence of PSemspecs associated withthe 
marker then further constrains the other sentence planning decisions (see figure 3). 
Now, it might also be the case that the lexicalization expert (e.g., MOOSE) calls for verbalizing 
the result of the cooling process (proposition \[15\]) only and proposes the lexeme be cool Now, the 
marker after is out, as it requires the satellite to be realized as a subordinate clause with a process 
of type activity (see lexicon entry of AFTER in figure 3). Alternatively, if the lexeme chosen is 
cool down, markers such as with are not available, as its PSemSpec allows for combining with a 
property-ascription only. Now, if some other expert decides to use a negation with the nucleus, 
unless is selected as marker since it expects a negativ e polarity in the nucleus; its (partial) lexicon 
entry is shown figure 3. 
Selecting unless in turn restricts the options for other sentence planning tasks, since its PSem- 
Spec states that a hypotactic structure with the subordinate clause in sentence-final position is 
needed (due to the :theme X line). In short, decisions can be propagated in both directions: from 
other formative decisions to marker choice, and from marker choice to other decisions. Imagine 
that the process of tree linearization be driven• by the overall goal of producing concise text; in 
this case, the flexibility in ordering decisions allows for producing short text by choosing with and 
letting the other decisions follow. 
We have characterized a constraint-based mechanism that does not impose a strict order on 
making decisions in linearizing the discourse tree. Various ways of implementing such a scheme can 
be imagined; one is the blackboard-based approach suggested by Wanner and Hovy \[1996\], another 
is the "Hunter-Gatherer" search paradigm introduced by Beale \[1996\]. 
6 Summary 
Present-day text generation systems typically employ quite simplified approaches for signalling 
discourse relations in text. Our work aims at enabling generators to truly choose discourse markers 
on the basis of generation parameters and context. This way, we gain variety in marker usage 
that is not just random but controlled. Furthermore, we are interested in a uniform, declarative 
representation of the information necessary. Approaches that encode marker choice in the grammar 
(such as Vander Linden \[1994\]), while certainly an improvement over previous h l mappings between 
relations and markers, loose flexibility when it comes to account for the interactions between marker 
choice and other sentence planning decisions. 
These considerations led us to develop a lexicon of discourse markers. While its construction 
is still in progress, we have shown samples of the kind of representations we are aiming at, and we 
have demonstrated how such lexicon entries can be employed as a resource in the sentence planning 
phase of the generation process. In our framework, lexicon entries consist of applicability conditions 
(for deciding whether the marker is a candidate at all), partial sentence semantic specifications (for 
combining the marker with neighbouring constituents), and distinguishing features for paradigmatic 
choice. We have described how an input discourse structure tree can be linearized into a sequence 
136 
I 
I 
| 
I 
I 
I 
ii 
I 
I 
I 
il 
:li 
I 
i I 
of sentence plans, given a sentence planner that exploits the information supplied by a discourse 
marker lexicon. 

References 
\[Bateman 1997\] Bateman, J. 1997. Enabling Technology for Multilingual Natural Language Generation: The KPML 
Development Environment. In Journal of Natural Language Engineering, 3(1), 15-55. 
\[Beale 1996\] Beale, S. 1996. Hunter-Gatherer: Applying Constraint Satisfaction, Branch-and-Bound and Solution 
Synthesis to Computational Semantics. NMSU Technical Report, MCCS-96-289. 
\[Delia et al. 1996\] Delin, J., D. Scott, A. Hartley. 1996. Pragmatic Congruence through Language-Specific Mappings 
from Semantics to Syntax. In Proc. of the 16th Conference on Computational Linguistics, Copenhagen, 292-297. 
\[DiEugenio et al. 1997\] DiEugenio, B., J. Moore, M. Paolucci. 1997. Learning Features that Predict Cue Usage. In 
Proc. of the 35th Conference of the Association for Computational Linguistics, Madrid, 80-87. 
\[Dorr and Gaasterland 1995\] Dorr, B., T. Gaasterland. 1995. Selecting Tense, Aspect and Connecting Words in 
Language Generation. In Proc. of the l$th International Joint Conference on Artificial Intelligence, Montreal, 
1299-1305. 
\[Elhadad and McKeown 1990\] Elhadad, M., K.R. McKeown. 1990. Generating Connectives. In Proc. of the 13th 
Conference on Computational Linguistics, Helsinki, 97-101. 
\[Grote et al. 1997\] Grote, B., N. Lenke, M. Stede. 1997. Ma(r)king Concessions in English and German. In Discourse 
Processes 24(1), 87-118. • • 
\[Halliday and Hasan 1976\] Halliday, M:A.K., R. Hasan. 1976. Cohesion in English. London: Longman. 
\[Helbig and Buscha 1991\] Helbig, G., J. Buscha. 1990. Deutsche Grammatik: FAn Handbuch fiir den Ausiiinder- 
unterricht. Berlin, Leipzig: Langenscheidt, Verlag Enzyklop~die. 
\[Kasper 1989\] Kazper, R. 1989. A Flexible iInterface for Linking Applications to Penman's Sentence Generator. In 
Proc. of the DARPA Workshop on Speech and Natural Language Processing, University of Pennsylvania, 153-158. 
\[Knott and Mellish 1996\] Knott, A., C. Mellish. 1996. A Feature-based Account of the Relations Signalled by Sen- 
tence and Clause Connectives. In Language and Speech 39 (2-3). 
\[MacGregor and Bates 1987\] MacGregor, R., R. Bates. 1987. The Loom Knowledge Representation Language. Tech- 
nical Report ISI/RS-87-188, USC/ISI. 
\[Mann and Thompso n 1988\] Mann, W., S. Thompson. 1988. Rhetorical Structure Theoryl Towards a functional 
theory of text organization. In TEXT, 8(2), 243-281. 
\[Martin 1992\] Martin, J. 1992. English Text - System and Structure. Amsterdam: John Benjamins. 
\[Moser and Moore 1995\] Moser, M., J. Moore. 1995. Using Discourse Analysis and Automatic Text Generation to 
Study the Discourse Cue Usage. In Proc. of the AAAI Spring Symposium on Empirical Methods in Discourse 
Interpretation and Organization. 
\[Nirenburg et al. 1989\] Nirenburg, S., V. Lesser, E. Nyberg. 1989. Controlling a Language Generation Planner. In 
Proc. of the 11th International Joint Conference on Artificia! Intelligence, Detroit. 
\[Panaget 1994\]' Panaget, F. 1994. Using a Texual Representation Level Component in the Context of Discourse and 
Dialogue Generation. Proc. of the 7th International Workshop on Natural Language Generation, Kennebunkport, 
Maine, 127-136. 
\[Quirk et al. 1972\] Quirk, R., S. Greenbanm, G. Leech, J. Svaxtvik. 1992. A Grammar of Contemporary English. 
Harlow: Long'man, (20th ed). 
\[Rainbow and Korelsky 1992\] Rainbow, O., T. Korelsky. 1992. Applied Text Generation. In Proc. of the C6nference 
on Applied Natural Language Processing, Trento. 
\[RSsner and Stede 1992\] RSsner, D., M. Stede. 1992. Customizing RST for the Automatic Production of Technical 
Manuals. In R. Dale et al. (eds.) Aspects of Automated Natural Language Generation. Berlin: Springer, 199-214. 
\[Scott and de Souza 1990\] Scott, D., C. de Souza. 1990. Getting the Message across in RST-based Text Generation. 
In R. Dale et al: (eds.) Current Research in Natural Language Generation. Academic Press, 31-56. 
\[Stede i996\] Stede, M. 1996. Lexical Paraphrases in Multilingual Sentence Generation. InMachine Translation 11, 
75-107. 
\[Vander Linden 1994\] Vander Linden, K. 1994. Generating Precondition Expressions in Instructional Text. In Proc. 
of the 15th Conference on Computational Linguistics, Kyoto. 
\[Wanner and Hovy 1996\] Wanner, L., E. Hovy. 1996. The HealthDoc Sentence Planner. In Proc. of the 8th Interna- 
tional Workshop on Natural Language Generation, Herstmonceux Castle, 1-10. 
