Planning Word-order Dependent Focu s Assignments* 
Cornelia Endriss and Ralf Klabunde 
University of Heidelberg 
Center for Computational Linguistics 
Karlstr. 2, 69117 Heidelberg, Germany 
{endriss, klabunde}@j anus. gs. uni-heidelberg, de 
Abstract 
Word order and accent placement are the primary 
linguistic means to indicate focus/background struc- 
tures in German. This paper presents a pipelined ar- 
chitecture for the generation of German monologues 
with contextually appropriate word order and accent 
placements for the realization of focus/background 
structures. Our emphasis is on the sentence plan- 
ner that extends the respective propositional con- 
tents with discourse-relational features and decides 
which part will be focused. Such an enriched se- 
mantic input for an HPSG-based formulator allows 
word order variations and the placement of prenu- 
cleus and nucleus accents. Word order is realized 
by grammatical competition based on linear prece- 
dence (LP) rules which are based on the discourse- 
relational features. Accent placement is realized by 
a syntax-driven focus principle that determines the 
focus exponent and possible bearers of prenucleus 
accents within the syntactically realized focus, the 
so-called focus domain. 
1 Focus and word order 
determination as sentence 
planning tasks 
This paper addresses aspects of the control of in- 
tonation belonging to the area of sentence planning 
\[Beale et al., 1998: Wanner and Hovy, 1996\]. In 
many s, intonation can reflect pragmati- 
cally motivated conceptual decisions. In particu- 
lar, focus/background structures (FBSs) reflect the 
speaker's beliefs of the listener's information state. 
Since FBSs are realized in German primarily by 
word order dependent accent placements, focus plan- 
ning and word order determination are subtasks of 
sentence planning. 
Due to the complex interactions anaong the vari- 
ous subtasks of sentence planning \[Hovy and Wan- 
* The research reported in this paper is funded by the DFG 
(German Science Foundation) in the priority program "Lan- 
guage Production" under grant lie 146T/3-I. The authors 
would like to thank the three anonymous referees for helpful 
comments and suggestions. All remaining mistakes are, of 
course. OUF o~vn. 
her, 1996\] proposed a blackboard-based sentence 
planner instead of a pipelined architecture. How- 
ever, we will demonstrate - as a byproduct of our 
approach to focus planning - that in some cases the 
complexity of interactions can be realized by a tradi- 
tional top-down expansion process. The intertwined 
clause-internal organization of focus planning and 
word order determination for the realization of FBSs 
in German is obtainable by hierarchical planning. 
In what follows, we will first give examples of the 
interplay of focus and word order in German. We 
will present the architecture of our NLG system that 
realizes FBSs and describe in more detail hierarchi- 
cal sentence planning for FBSs. By means of some 
examples we are finally showing how word order de- 
pendent focus assignment works. 
2 The interplay of focus and word 
order 
The pragmatic function of the FBS is to indicate 
to the listener of an utterance that a certain part 
of that utterance has been put into the foreground. 
The semantic information of this foregrounded part. 
has either been selected from a set of alternative 
beliefs ascribed t.o the listener, or it is a revision 
of certain beliefs (in case of contrastive focus), or 
the focused phrase expresses 'new' information the 
listener does not know or is not able to infer from 
his beliefs \[Halliday, 1967\]. 1 
In all three cases the focus domain - the syntact i," 
realization of a focus - contains the so-called focus 
exponent, i.e. the bearer of the focal accent which 
we are identifying with the obligatory' nucleus ac- 
cent. The existence of this accent indicates to the 
listener that one part of the message conveys one 
of these three functions. In addition to the nucleus 
accent optional prenucleus accents can exist as well 
which do not have a discourse function in general. 
I This list of apparently diverse functions shows that there 
are possibly several phenomena which have been labeh!d as 
'focus' within the last 70 years or so. There is an ongoing 
discussion in the linguistic community whether these three 
functions can be traced back to one common principle (of. 
\[Schwarzschild, 1999\]). 
156 
but a prosodic function which goes back to diverse 
planning phenomena ......... ~. 
Three examples shall demonstrate the interplay of 
word order with accent placement. Example (3) is 
from our speech corpus of retellings of a trick film we 
analyzed to obtain rules for accent placement. The 
other examples are variations of (3) showing that dif- 
ferent word order and accent placement correspond 
to different focus domains. For reasons of simplicity 
we are abstracting from specific pitch accents in this 
paper. Accent bearers are given, in capitals. Fur: 
thermore, tile examples do not exhibit prenucleus 
accents. However, our rules for accent placement 
account for these accents as well. The reason is that 
prenucleus accents are determinable if the bearer of 
the nucleus accent is known. 
1. er fiillt in die STEINebene runter 
he falls in the stone plateau down 
'he is falling down to the stone plateau' 
2. in die STEINebene fiillt er runter 
3. er f6llt RUNter in die STEINebene 
Semanticians pointed out that the key concept for 
word order and its consequences for accent place- 
ment is contextual boundedness (see, e.g., \[Jackend- 
off, 1972; Rooth, 1992\]). However, their method 
to simulate the different contexts by questions the 
sentence is able to answer tells us only something 
about the number of possible loci. For example, 
sentence (1) is able to answer five possible questions, 
depending on which constituent provides the answer 
(the contexts range from Which specific plateau is he 
falling down to? with focus on tile compound only to 
What's up?, focusing the whole sentence). Example 
(2), however, with the locative PP in sentence-initial 
position but identical accent placement is only able 
to answer three questions. Hence, (2) is contextu- 
ally more bounded than (1). Example (3) exhibiting 
an extraposed unit clearly demonstrates the need for 
an FBS-related word order. Extrapositions are the 
linguistic means in German to separate sense units. 
Tile extraposition is used to mark two informational 
units: first, the person is falling down and second 
that the resulting place is the stone plateau. Since 
informational units coincide with prosodic phrases, 
each phrase contains one nucleus accent so ti~at, two 
separate focus domains exist. 
From an NLG perspective explaining word order 
and accent placement by tile possibility to answer 
context questions points to tile wrong direction. Nei- 
ther should we generate isolated sentences nor are 
we interested in focus ambiguities. Rather we have 
to determine a certain word order with a twofold 
purpose: first, it must be able to express a planned 
focus and second, it should guarantee coherence of 
~he text. 
To our knowledge, the problem of how word or- 
...... ...der`~m¼td.:fc"cM`dÜmain¼determin:`.~`ti~n``interact` has not 
been adressed in NLG research yet. The SYNPHON- 
ICS formulator \[Abb et al., 1995\] that is able to gen- 
erate German single sentences.with FBSs does not 
take into account the interplay between word order 
and accent placement. Instead word order is deter- 
mined by incremental syntactic construction; situa- 
rive factors have not been addressed in this system. 
The SPEAK! system \[Teich et al., 1997\] also does not 
account_for the interplay.of word .order with. accent 
placement. However, this system cannot be directly 
compared with our approach, since the coverage of 
phonological phenomena is completely different: We 
are interested in FBSs in monologues, whereas the 
SPEAK! system primarily accounts for the role of a 
dialogue history to achieve the assignment of various 
intonation patterns. 
Generally, the realized word order of an utter- 
ance is the result of its embedding into the situa- 
tive context, which finds expression in the use of 
linear precedence (LP) rules for word order determi- 
nation during surface realization. The idea is that 
constituents are ordered with respect to preferential 
properties expressed by these LP-rules. From an 
NLG perspective the question is, then, where the 
information comes from that allows us to make use 
of these LP-rules?" In our approach we derive the 
information necessary for the use of LP-rules from 
a discourse model that relates various aspects of a 
discourse to one another. Since we are generating 
monologues only the utterances previously produced 
by the program require consideration. 
Tile generation of monologues with appropriate 
word order and focus/background structures com- 
prises five major tasks: 
1. The information to be conveyed must be se- 
lected and linearized by a content planner. 
2. During sentence planning: 
(a) foci n-lllSt be determined, and 
(b) conditions for word order realization must 
be given. 
3. During surface realization: 
(a) the loci must be mapped onto focus do- 
mains while tile sentences with their re- 
spective word order are fornmlated, and 
(b) the bearers of (pre)nucleus.accents within 
each focus domain must be determined. 
Since this paper addresses sentence planning, we are 
focusing on tasks (2a) and (21)) only. We are leav- 
ing aside content planning (task 1) because the lit> 
earization problem does not affect FBS determina- 
lion. The content planner provides the respectiw' 
propositions lhal will be extended during sentence 
157 
planning by pragmatic information for realizing the 
FBS. The result of sentence planning functions .as 
input for a competition-based fornmlator. In or- 
der to demonstrate how the formulator is able to 
realize FBSs by means of grammatical competition, 
we will also outline the determination of focus do- 
mains, word order, and accent bearers in focus do- 
mains (tasks 3a and 3b). 
3 Architecture of FOGS 
The five tasks mentioned above are realized in our 
NLG system FOGS. 2 Currently the system generates 
brief retellings era trick film with each sentence hav- 
ing a contextually appropriate word order and focus- 
relevant prosody with the context provided by the 
discourse model. Figure (1) shows the architecture 
of the system. Sentence planning takes into consid- 
eration the current state of a discourse model. When 
constructing the input for the formulator, the dis- 
course model will be continuously updated so that 
the word order of the currently planned sentence is 
coherent with the word order of the preceding sen- 
tence. Word order relevant information is encoded 
by discourse relational features of discourse refer- 
ents. 
The HPSG-based formulator realizing the sen- 
tences uses weighted LP-rules for word order de- 
termination that take into account the discourse- 
relational features in the semantic input. Bearers of 
(pre)nucleus accents within focus domains are deter- 
mined by a focus principle. 
4 Sentence planning in FOGS 
The planning operators creating the input for the 
formulator cause the transition to new states of the 
discourse model. The initial state of the discourse 
model is characterized by the lack of any information 
on the events to be conveyed. Correspondingly in 
the goal state all events are represented. 
Our discourse model is a knowledge store consist- 
ing of two major registers. It consists of a Discourse 
Representation Structure (DRS. cf. \[Kamp and 
Reyle, 1993\]) (R. I'() with sets of mutually known 
discourse referents R and DRS-conditions h, and a 
set Ref of referential movements assigned to the dis- 
course referents, lI/eferential movements det.ermine 
how discourse referents "are passed on from one sen- 
tence to tile next one. H is a pair (RA, RN} consist- 
ing of referents of tile directly preceding utterance 
and referents of all-other previous utterances. Since 
referential movements are typically linked with iden- 
tifiability conditions for discourse referents, ttle lat- 
ter can be derived front the former. New referents 
are declared as being unidentifiable for the listener. 
2Not to be confused with FOG, a system that generates 
weather forecasts\[Goldberg t t a/.. 199-I\]. Foc;s is the acronym 
for "focus generation system'. 
% \[ discourse relational q f~res refined plan q focu~ground \[ 
} deleTtion I 
pragmatically enriched 
input 
Lemma Selector 
lemmas 
Formulator \] 
Prosodically annotated 
monologue 
Conienl Planner 
linearizalion 
i abs~aci plan 
Sentence Planner 
I 
Figure 1: Architecture of FOGS 
while re-established ones should typically be identi- 
fiable by a definite description. Maintained referents 
are usually anaphorically identifiable. Furthermore, 
alternative sets All are determined by sortal restric- 
tions. Discourse referents function as alternatives if 
they are stored in the discourse model ill R and are 
instances of the same superordinated concept. Anal 
ogously, concepts are alternatives if they are stored 
in the discourse model in K and possess the same 
directly superordinated concept. During planning 
the discourse model will be continuously updated. 
Updating comprises the insertion of new discourse 
referents into RA. shifting referents from t¢.-I to I5.\. 
and, in case of referential re-establishment, shifting 
referents from RN to RA. Furthermore, new \[)FIS 
conditions will be introduced into )Y, and the ref- 
erential movement conditions are updated, resulting 
also in new alternative sets and identifiability con- 
ditions Id. 
W~ use a hierarchical planner \[Sacerdot.i, .1974\]. 
The content planner provides the abstract plan. 
Plan refinement during sentence planning con- 
sists of the proposition-wise introduction of oper- 
ators for the discourse relational features and fo- 
cus/background determination. The result of apply- 
ing the operators to tile single propositions functions 
as nput Io the formulator. 
158 
4.1 Discourse-relational features 4.2 Focus and background determination 
Extending the propositions by.-disc~otrrsecelationat ......... We~lreadypointed~out.t.h~t-.foGusing-asemanticrep-. 
features makes intensive use of the discourse model. 
Three discourse relational .factors influencing word 
order are realized as plan operators: topic assign- 
ment, referential movement, and identifiability of 
discourse referents by the listener. 
Topic assignment: Topics establish an 
aboutness-relation between a familiar discourse 
referent and the sentential predication. We adopt 
:the conditions for tooic:assignment.-:propoSed "in' 
\[Klabunde and Jansche, 1998\]. Topic candidates 
must be identifiable discourse referents and they 
should be as high on a so-called topic acceptance 
scale as possible. According to such a scale referents 
that are currently lit up constitute the best topic 
candidates. In our approach, these are referents 
from the intersection of RA and the referents of 
the current event proposition to be realized. The 
topic acceptance scale is mirrored in the successive 
application of operators for topic assignment. For 
example, if several referents as candidates exist, 
a discourse referent will be chosen that is marked 
as anaphorically identifiable and referentially 
maintained: 
topicAssignment (Event ,Eventl) : 
PKE: \[(RA f7 EventRs = Topics), 
E 6 Topics, 
anaphId(E) 6 Id, 
refMaintained(E) 6 Id\] 
EFF: \[topic(E) U Event = Eventl\] 
DEL : \[\] 
The LP-rule referring to topic assignment is topic 
< :focus, i.e. topics should precede loci. 
Referential movement: Referential movement 
comprises the picking up of discourse referents from 
previously uttered information and the introduc- 
tion of new referents, respectively. If referents 
from the directly preceding utterance are picked up, 
these referents are maintained. Referents from all 
other previous utterances are re-established. Flef- 
erential movement influences word order because 
maintained referents are usually realized before re- 
established ones, and re-established ones precede 
new referents, as indicated by the following LP-rule: 
refMaintained < refReEstablished < new 
Identifiability: With respect to identifiability 
of discourse referents, we distinguish between 
anaphoric identifiability, identifiabilit.y by a definite 
description, and referents that. are non-identifiable 
for the listener. Identifiability influences word or- 
der as well because anaphorically identifiable ref- 
erents are usually realized before dethfites and 
those precede non-identifiable referents: anaphld < 
de:finiteld < nonId. 
resentation is based on one of three functions: the 
selection of beliefs from a set of alternatives, con- 
trasting a belief with a different one, and indicating 
new information. These three functions have also 
been verified in our corpus of story tellings. Each of 
these functions has been treated separately in vari- 
ous systems (see, e.g., \[Prevost and Steedman, 1993\] 
for contrastive focus in a concept-to-speech system, 
.\[Thetme.:et _al.,..:1997.\]. for, n .ew :in.formation in a .data.-. 
to-speech system, and \[Blok and Eberle, 1999\] for 
alternative semantics in machine translation), but 
a single and comprehensive approach has not been 
proposed yet. However, structure and content of 
our discourse model allow us to determine FBSs by 
means of planning operators as well. Different pre- 
conditions for the focus determining operator result 
in the successive check whether one of these three 
functions is satisfied. First it is checked whether the 
proposition to be conveyed contains any informa- 
tion that is new for the listener. New information 
is what is not stored in the DRS K of the discourse 
model, a If these preconditions are not satisfied it is 
checked whether parts of the proposition belong to 
alternatives presumed by the listener. Only if these 
preconditions fail a contrasting focus is realizable. 
Contrasting focus is realized if some property in K 
of an activated discourse referent in RA contradicts 
a property in the semantic input under considera- 
tion, provided the same sortal restriction holds as 
for the alternatives. 
5 Surface realization as grammatical 
competition 
The resulting input for grammatical competition is 
a blend of semantic and pragmatic information. For 
example, the input for realizing example (1) is as 
follows: 
focus: \[fallingDowa(e,m), into(e,s), stonePlateau(s), 
definiteld(s), refReEstablishment(s)\] 
ground: \[man(m), anaphld(m), refMaintenance(m), 
topic(m)\] 
The constants m, s, and e are referents for a spe- 
cific man, stone plateau, and tile evenl of falling 
down. The values of the features :focus and ground 
represent the focused part of the proposition and 
• the background, respectively. While tile realization 
of the focus domain is the task of converting the 
complete focus into one phrase, word order will be 
determined by LP-rules that pick up the pragmati- 
call2,' motivated literals on topichood, identifial)ility, 
and referential movement. 
aThis implies that we are ignoring any inferential capabil- 
ities in the current system. 
159 
As already mentioned, the notion of grammati- 
cal competition is necessary to account, for the in-. 
teraction of syntactic and phonological constraints 
on focus/background structures. The idea to use 
a competition model to explain word order varia- 
tions in German is not new (of. \[Steinberger, 1994; 
Uszkoreit, 1987\]). The advantage of grammatical 
competition compared to a non-competitive use of 
precedence rules (as in standard HPSG) is its flex- 
ibility. A competition model allows to take syntac- 
tic as well as semantic and pragmatic preferences 
into consideration, and to determine'the acceptabil- 
ity of a sentence with respect to the situative con- 
text. The usual approach is to formulate preference 
rules which have a certain impact on the naturalness 
of constituent orders. Some of these preference rules 
are stronger than others. The number of preference 
rules which are satisfied or violated, in combination 
with the relative importance of the different factors, 
is responsible for the varying degree of naturalness 
of word order variations. Analogously to this idea 
we use weighted LP-rules as well which are based 
on the planned discourse-relational features. 
Focus domains are realized by means of a focus 
principle. Applying the focus principle results in 
the projection of a focus feature to the dominating 
node. Together with the standard HPSG-principles 
the focus principle confines the successive applica- 
tion of the head-complement, head-filler, and head- 
adjunct schemata to two lemmas in order to build 
up phrases and sentences. The focus principle con- 
strains the placement, of prenucleus and nucleus ac- 
cents in view of the syntactic status of the phrasal 
signs. It is based on the following empirically vali- 
dated regularities with respect to the placement of 
the nucleus and prenucleus accents: 
1. in phrases with a head-daughter and adjunct- 
daughter the focus exponent is in the head- 
daughter and a prenucleus accent is in the 
adj unct-daughter. 
2. for phrases with a head-daughter and 
complement-daughter holds: 
(a) if the head-daughter is a verbal projec- 
tion, the focus exponent is in the head- 
daughter and a prenucleus accent is in the 
complement daughter. 
(b) else the accents are in the complement-- 
daughter. 
The regularities underlying the nucleus and prenu- 
cleus accent placement have been formulated on the 
basis of an analysis of a story telling corpus. The 
tellings have been analyzed w.r.t, the position of 
pitch accents and their indication of possible fo- 
cus domains. Two resuhs of \[his analysis shall he 
memioned here: First, the analysis showed that the 
overwhelming number of focus domain determina- 
gEon can.be explained.by.syntax-based.projection 
rules (see, e.g., \[Gfinther, 1999; Ladd, 1996\] for some 
proposals) underlying our focus principle. Second, 
given the three basic pragmatic functions of FBSs, 
primarily information that was new to the listener 
has been accented. Contrastiveness was confned to 
focal accents on certain closed-class items such as 
determiners. 4 
While focus domains are realized by a syntactic 
principle, word order will be realized by means of 
weighted-LP-rules. Since especially the LP-rule 
• topic < focus requires information on focused con- 
stituents focus determination must be completed be- 
fore word order will be realized. We introduced the 
necessary LP-rules in section 4.1. 
Based on these LP-rules word order will be deter- 
mined by means of the operation of domain union 
proposed in \[Reape, 1994\]. If the head or the daugh- 
ter is a verbal projection the domain of the phrase 
will be received by domain union. Verbal projec- 
tions are of interest for word order realization be- 
cause only in this case the LP-rules will be evalu- 
ated. Otherwise the domains will be combined ac- 
cording to the directionality feature DIR of the head 
and a MOD-DIR feature of an adjunct. The for- 
mer determines the order of head and complement, 
while the latter is responsible for the order of ad- 
juncts and their modified element. Since in this case 
no LP-rules have to be evaluated, word order deter- 
ruination is a trivial task. 
6 Results 
The system just described produces brief retellings 
of one episode of the aforementioned trick film based 
on a knowledge base representing the single events 
and a discourse model. Depending on the content of 
the discourse model word order of the respective sen- 
tences and focus assignments differ. We are giving 
one detailed example showing the different status of 
the discourse model and its influence on the realiza- 
tion of word order and the FBS. After that some 
texts FOGS is able to generate are presented. 
Let the content of the discourse model be as fol- 
lows: 
RAo = @ 
RNo = {s,m,d,...} 
A'O = {desertPlateau(d), stonePlateau(s), 
littleMan(m), fallingDowll(e,m), ...} 
Refo = {refMaintenance(s), 
refReEstablishme~t(m),...} 
Alto = {alt(stonePlatea,(s), {desertPlateau(d)}), 
alt(in(e,s),{on(e,s)})} 
Id 0 = {definiteld(m), definiteld(s),...} 
.1Note that determiners as bearers of nucleus accents do 
not constitute a problem for our system. In this case only 
the identifiability condition belongs to the focus, which will 
I.~,. mapped ont,~ a corresponding lemma. 
160 
The content planner deterlnines 
fallingDown(e,.m) ,-in(e,.s), s~oneP&ateau(-s) ,. 
littleHan(m) as the proposition to be conveyed. 
The operator for topic assignment marks s as 
sentence topic because it is the currently best 
available topic according to the topic acceptance 
scale. Referential movement is as follows: since s 
has been declared in the discourse model as being 
referentially maintained, it will be maintained in the 
first utterance as well. Discourse referent m was re- 
established and, therefore, will become referentially 
maintained. Since referent S wasidentifiable by /i 
definite description for the listener and is the topic, 
it remains identifiable by definite means, resulting in 
a definite NP. Referent m was identifiable by definite 
means and becomes anaphorically identifiable. 
Focus and background are determined as follows: 
first it is checked whether any information in the 
proposition is new to the listener. Since all liter- 
als from the propositional content also exist in if, 
nothing can be focused due to being 'new' informa- 
tion. However, there are two literals in the propo- 
sitional content with explicitly represented alterna- 
tives. Since both literals can be linguistically real- 
ized as one constituent (as a PP), only one focus 
domain and one focus exponent will appear. 
The resulting utterance is in die STEINebene fa'llt 
er runter with the locative PP in clause-initial posi- 
tion and accent on the noun-noun compound. The 
PP has been fronted because it receives the best eval- 
uation w.r.t, the three discourse-relational features. 
The compound functions as focus exponent because 
the whole PP constitutes the focus domain, which is 
managed by the focus principle. 
Taken all effects of the planning operators to- 
gether, the updated discourse model is as follows: 
R~, = {~,~} 
RNI = {d .... } hl = No 
Refl = {refReEstabl ished(s) , 
refMaintenoalce (m) , . . . } 
Ale1 = Alto 
Idl = {anaphId(m), de:finiteId(s),...} 
The next proposition to be conveyed is 
i itt leman (m), walk ingAround (e i, m). T he 
new senterlce topic will be m. Anaphoric identi- 
fiability and referential maintenance of referent ra 
would usually result in keeping these .conditions 
for the linguistic realization so that a pronoun will 
be generated. However, the topic shift results in 
a change to identifiability by a definite description 
for m in the following sentence. The event literal 
is focused because it provides new information, 
resulting in the VP as focus domain. According to 
the focus principle, the prefix receives the nucleus 
accent. 5 The resulling utterance is Das Ma',mchen 
:; \Ve adopl an approach to the goneration ,~f separable pre- 
fix verbs proposed by \[Le\[,+'t h. 1992\] that is based on adjunc- 
161 
Muff umHER (the little man walks aROUND). 
Exemplary brief retellings of the, film.epi~ode:gen- .. 
erated by FOGS are given below. Note that the single 
sentences always express the same respective propo- 
sitional content. What ..differs are the word order 
and/or the position of the nucleus accent. These dif- 
ferences in word order and accent placement are due 
to the varying content of the discourse model, the 
different possibilities for focus determination, and 
the effects of the applicable planning operators for 
referential movement, topic assignment, and identi- .. . • __., . , . . 
fiability. " 
1. In die Steinebene fa'llt es runter auf der Suche 
nach Wasser. 
Das Miinnnchen Muff umHER. 
PlStzlich schieflen STEINttirme aus der Erde 
auf. 
Die Steinttirme erheben das MANNchen. 
A rough translation is: 
While looking for water it is falling down to 
the STONE plateau. The little man is running 
aROUND. Suddenly stone pillars are shooting 
up from the GROUND. The stone pillars lift up 
the little MAN. 
2. Es fiillt auf der Suche nach Wasser in die 
" STEINebene runter. 
Es hluft umHER. 
PlStzlich schieflen aus der ERde Steintiirme a@ 
Die Steintiirme erheben das Mii'NNchen. 
3. Es fiillt auf der Suche nach Wasser m die 
STEINebene runter. 
Es lh'uft umHER. 
PlStzlich schieflen STEINtiirme aus der Erde 
auf. 
Die Steint(irme erheben das M,4"NNchen. 
4. Es fSllt auf der Suche nach WASser m die 
Steinebene runter. 
Es 15uft umHER. 
Pldtzlich schieflen STEINt(irme aus dcr l\]rdc 
a uf. 
Dic Stemtiirme erhebeu das ,U:i'N:Vche, 
7 Summary and outlook 
Sentence planning for the realization of fo- 
cus/background structures in German comprises the 
-determination of .discourse relational.features for 
the realization of an appropriate word order and 
the determination of the focus of the respective 
propositions. With this information available a 
competition based formulator is able to realize the 
focus donlain, word order, and accent bearers within 
the focus domain. 
tion and compatil)le wit h our fi)cus principle. 
While hierarchical planning of discourse relational 
features for the use of LP-rules during surface real- 
ization seems to be a promising approach for word 
order dependent focus/background determination, 
additional constraints on word order are required 
to block overgeneration. The three LP-rules 6 are 
necessary, but no sufficient means to determine free 
word order in general. Next to these discourse- 
related rules there are syntactic and semantic re- 
strictions on word order as well. However, due to 
their discourse-related nature, the conditions for us- 
ing our LP-rutes can be-derived from the. discourse 
model. Syntactic and semantic constraints on word 
order do not require a discourse model. 
To summarize, the combination of hierarchical 
content planning with grammatical competition in 
view of a focus principle seems to be a promising 
approach for focus/background determination. 

References 
Abb, B; G/inther, C.; Herweg, M.; Lebeth, K.; 
Maienborn, C. and Schopp, A. (1995) Incremental 
syntactic and phonological encoding - an outline 
of the SVNPHONICS-Formulator. In: G. Adorni 
and M. Zock (Eds.) Trends in Natural Language 
Generation. Berlin: Springer. 
Beale, S.; Nirenburg, S.; Viegas, E. & Wanner, 
L. (1998) De-Constraining Text Generation. Pro- 
ceedings of the Ninth International Workshop 
on Natural Language Generation. Niagara-on-the- 
Lake, Canada; 48-57. 
Blok, P. & K. Eberle (1999) What Is the Alter- 
native? The Computation of Focus Alternatives 
from Lexical and Sortal Information. In: P. Bosch 
& R. van der Sandt (Eds.) Focus. Linguistic, 
Cognitive, and Computational Perspectives. Cam- 
bridge: Cambridge University Press; 105-120. 
Goldberg, E.; Driedger, N. & Kittredge, R. (19!)4) 
Using Natural-Language Processing to Produce 
Weather Forecasts. IEEE Expert, 9 (2): 45-53. 
G/inther, C. (1999) Prosodic und gpraehproduktzon 
T~bingen: Niemeyer. 
ttalliday. M.A.K. (1967) Notes on transitivily and 
theme in English. Journal of Lmguzstics. 3: 199- 
244. 
Hovy, E. & L. Wanner (1996) Managing Sentence 
Planning Requirements. Proceedings of the ECAI 
Workshop Gaps and Bridges: New Directions in 
Planning and Natural Language Generation. 
Jackendoff, R. (1972) Semantic .Intewretation tn 
Generative Grammar. Cambridge, Mass.: MIT 
Press. 
Kamp. t1. & U. I~eyle (1993) From Discourse to 
Logic. I)ordrecht: l(luwer. 
6topic < focus: refMaintained < refRegstablished < 
new: anaphId < definiteId < nonId 
Klabunde, R. & M. Jansche (1998) Abductive rea- 
soning for syntactic realizatiot|:.Proceedings of the. 
Ninth International Workshop on Natural Lan- 
guage Generation. Niagara-on-the-Lake, Canada; 
108-117. -~ 
Ladd, D.R. (1996) Intonational Phonology. Cam- 
bridge: Cambridge University Press. 
Lebeth, K. (1992) Zur Analyse yon trennbaren 
lokalen Prdfizverben in der HPSG. Stuttgart, IBM 
Germany, IWBS Report 224. 
Prevost, S.. & M. Steedman (1993) Generating Con- 
textually Appropriate Intonation. Proceedings of 
the Sixth Conference of the European Chapter 
of the Association for Computational Linguistics. 
Utrecht; 332-340. 
Reape, M. (1994) Domain Union and Word Order 
Variation in German. In: J. Nerbonne, K, Net- 
ter & C. Pollard (eds.) German in Head-Driven 
Phrase Structure Grammar. CSLI Lecture Notes; 
46. Stanford, CA: CSLI; 151-197. 
Rooth, M. (1992) A Theory of Focus Interpretation. 
Natural Language Semantics I; 75-116. 
Sacerdoti, E. (1974) Planning in a hierarchy of ab- 
straction spaces. Artificial Intelligence, 5; 115- 
135. 
Schwarzschild, R. (1999) GIVENness, AvoidF and 
other Constraints on the Placement of Focus. To 
appear in: Natural Language Semantics. 
Steinberger, R. (1994) Treating 'free word order' 
in German. Proceedings of the 15th International 
Conference on Computational Linguistics (COL- 
ING). Kyoto. 
Teich, E.; E. Hagen; B. Grote & J. Bateman (1997) 
From comnmnicative context to speech: Integrat- 
ing dialogue processing, speech production and 
natural  generation. Speech Commumca- 
tion, ~1; 73-99. 
Theune, M.; E. Klabbers; J. Odijk ~ J.R. de Pi- 
jper (1997) Computing prosodic properties in a 
data-to-speech system. Proceedings of the work- 
shop on Concept-to-Speech Generation 5'.t/stems. 
ACL/EACL, Madrid; 39-45. 
Uszkoreit, tt. (1987) Word Order and Constttucnt 
Structure m German. CSLI Lecture Notes. 
Chicago: University of Chicago Press. 
Wanner, L. & E. How (1996) The HeahhDoc Sen- 
tence Planner. Proceedings of the Sth lntcrna- 
tional Workshop on Natural Language Genera- 
twn. Herstmonceux Castle, 1-10. 
