An HPSG-Based Generator for German 
An Experiment in the Reusability of Linguistic Resources 
Johannes Matiasek and Harald Trost 
Austrbm l/,esea, rch Institute 
for ArtificiM Intelligence* 
Schottengasse 3, 
A-1010 Vienna,, Austria, 
lBmaih {j ohn, hara\].d}0a±, un±vie, ac. at 
Abstract 
We describe the development of a gen- 
erator for (\]erman built by reusing and 
adapting existing linguistic data and 
software. Reusability is crucial for the 
successful application of N I,P techniques 
to real-life problems since it helps to cut 
down on both development and adap- 
tation effort. Itowever, combining re- 
sources not designed to work together 
is not trivial. We describe the prob- 
lems arising when integrating three pre- 
existing resources (FUF, a unification- 
based generator, an HPS(~ (\]ramrnm' 
for (~errnan, and X2MorF, 7~ two-level 
morphology (;omponent) and the adap- 
tat, ions necessary to come up with a wide 
coverage l;acticM generator for (\]erlnnn. 
1 Introduction 
A main obstacle for the successfifl application of 
N I,P is the necessary effort in terms of deve|op- 
inent and adaptation time. One possible answer 
to this problem is the use of generic and modu- 
la.r software. An example fbr a software system 
develot)e.d with this goal in mind is the FUF gen- 
erator (l~lhadnd, 1991), a well-documented pub- 
lic domain software written in IASI: ). Still, it is 
no straightforward task to employ that, kind of 
software for new applications. Another important 
step is the declarative definition of linguistic data 
(grammar and lexicon) which also fa(:ilit~tes reuse 
in another setting. The reuse of existing resources 
does not only save efforts but, to a hopeflflly much 
minor extent, also creates new tasks to be solved, 
i.e. the integration of resources not having been 
*The work reported here h,~s been carried out 
within tit('. I;t/E Project U/ST (LRM 062-09) and 
funded by the Austrian t;brschungsfdrderungsfonds 
dcr (lcwcrblichcn Wirtsch@, Grant 2/329. l"inan- 
(tim support for the Austrian Research Institute for 
ArtiticiM \[ntelligence is provided by the Austrian 
13undesministerium fiir Wisscnschaft, Forschung und 
Kunst. 
designed to work together, llow this can be done 
in an organized way is the topic of this paper. 
The work being described here was done in t, he 
context of a multilinguM text generation system. 
One Of the ohjectives of the project is to rouse ex-- 
isting resources for those subtasks for which at> 
propriate resources exist, l,'or the German tactical 
generator 1 an irnph;mentation of an HPS(, '~ style 
gramrm~r of (\]erman (used for parsing and genre> 
ation, but on a different software platform) was 
available inhouse. ~l)he I"UI ? generator was chosen 
as the core component of t.he system. 
However, two problems had to be solved before 
FUI" could be used for the planned purpose. One 
was the Net that li'UF, dew, loped for English, has 
no suitM)le morphologicM conlponent for the rich 
inflection of German. X2MorF (Trost, 1991), an 
avMlahle morphologicM component had go be iH 
tegrated with the FUI e generator for this purpose. 
The other probh;m was that the existing HPS(L 
inspired gramrna.r of German could not be directly 
ported to the FUF formMisln. 
Before we describe the integration task we will 
briefly sketch the mMn characteristics of these 
resources, emphasizing those aspects which ei- 
ther cause problems lbr integration or provide the 
means for perfi)rrning the integration task. 
2 Available Resources 
2,1 The FUF Generator 
li'UF (Elhadad, 1991) is a surface generator for 
naturM language based on the theory of flmctional 
unification grammar (Kay, 1979). It employs both 
phrase structure rules and unification of feature 
descriptions. Input to I"U F is a partially speciiied 
feature description which constrains the utterance 
to be generated. Output is a flflly specified fea- 
ture description subsumed by the input struel,ure, 
which is then linearized to yiehl a sentence. 
1 Tit(: task of a tactical generator is to produce sen- 
tentbd or subsentential phrases corresponding to a se- 
mantic input specitication and does not include text 
planning. 
2 Head Driven Phrase Structure Grammar (Pollard 
and Sa.g, 1987; Pollard arid S~tg, 1994) 
752 
2.1.1 Grammar Specification in FUF 
(~rammar and h;xi('.on are specified as one large 
\['eatm'e descril>Cion, containing aC h;asC one dis- 
.junction (given t>y Che alt keywor<l) ranging over 
the pllrasal and lexical cal;egories of l;he gr;umnar. 
The feature, cat is used Co indica/,c t, hese cat, e- 
Aeries. 'Fhe fealmre lex associat.es strings wil;h 
lexi<'at (;m;egories. '\['he trivial grammar of Fig. I 
exemplifies the layout of a FU F grammar. 
(alt (; --- S (with subject/verb agreement) 
((cat s) 
(subj ((cat rip))) 
(pred ((cat vp) 
(agr (^ - subj agr})))) 
--- NP (only proper nounu) 
((:at np) 
(n ((cat noun) (proper y)))) 
--- VP (only intransitive verbs) 
(cat vp) 
(v ((cat verb) 
(agr {^ " agr})))) 
; --- Lexicon 
((cat verb)(lex "laughs") 
(agr ((peru 3rd)(num ug)))) 
(((:at noun)(lex "Mary") 
(agr ((pets 3rd)(num sg))) 
(proper y)))) 
l,'igure l: A trivial FUF grammar 
}'oinCers are used Co enforce sCl'llcl;tlre sharing 
and provide a ~neans {,<> percolaCe informaCion 
within a \[eaCm'e sl;rlleCllre. 
I,'ll\],' provides the means Co specify a sul)sump 
don or(Jeriug of l?lpeS, whi<'h is useful {.<) express 
gem;ralizaLions, an(\[ it illa(:ro 1He('halliSlll. 
2.1.2 Operational Characteristics 
C, eneraCion sCari;s fl'om an undersl>e('ified input 
fe=al, urc struc/,ure, li'UF unifies the grammar inlo 
the input sCrucl, ure, i.e. enricln:s and furl.her con- 
strains iC. Alternatives are explored sequentially 
until one I>ranch succeeds. Thus Cite inl>ut st;rue- 
Cure never ('.ontMns all@motions. 
When unilieaCion aC the <:urren(, level is com- 
\[)leCe, i.e. noChing furl, her can I)e added to the 
int>ut sl, rucCure, every substructure of the input 
rei)resenCing acat, egory is recursively unified wil;h 
I;he gral~H\[nar. This process is re\[>eaCed breadth 
tirst mltil all <-onsCituents are h'.aves. 
To det, erinine whi<:h substructures have. to be 
processed rceursively FUF employs two methods. 
The defaull; strategy (;ollecCs all subsi, rucl, ures of 
Chc current level having a cat feature, l';xplicit 
specification o\[" subconsf, il;uents is also l>ossible via 
the special Feai.ure cset (<:onsCiCuent set). If cuet 
is presenl,, FUF performs recursion on these ex- 
l)liciCly given subsCruct, ur<'s only. E.g., \[,he de- 
fault sCraCegy operates on (;ategory s in Fig. 1 
as if (cset (subj pred)) had |teen specitied. 
When specifying (cset (pred)) only, no recur- 
sion wouht be performed <)n subj. 
2.1.3 Linearization 
The recursive unification process handles only 
t,he dominance relations of 1%c grannuar. In or- 
der to a.c(:ounl; for linear ordering of l, tlo resulting 
I,ree shaped feature structure, FUF performs a lin- 
earizal;ion \])l'OC(~S8 a~0,e.t" unification has linishcd. 
IAnear l)rac.e(hmc.c of ('.onstituenCs is sl)ecitied in 
the. grammar using (;he special feal;ure pattern. 
Only constituents mentioned in a pattern are real- 
ize(\[ during linearizalfion. Thus, the simple gram- 
mar in Fig. l has Co I)e enriched: (pattern (subj 
pred)) has Co be a.dded at (cat st, (pattern 
In)) has 1,o he ad(le(\] al~ (cat apt and (pattern 
Iv)) is needed aC (cat vp). I,exic.al categories 
don't, need a \[)att, crn feature. 
I>atterns need not spe(:ify an absolut, e ordering. 
E.g., (...a ...b ... ) q)eci\[ies t,haI, consCii;uenl, a 
Ires Co precede b. More sa(:h Imrl, ial pal,t, erns ntay 
Ice specified, i)aCtern unifical, ion leads Co all \]egM 
constiCuenl, (:ombinal, ions. 
I,inearizaCion traverses the ere(% ex|,ra,(:Cs the 
M,rings foun(l in Che lex \['eaCnre o\[" the leaves, and 
\[lai;Cens gl,is sCruclure a.ccording Co (;110 pattern 
directives fen u(l. 
2.2 The HPSG Grammar for German 
In IIPSG (Pollard and Sag, 1987; I)ollard and Sag, 
1994), i,he fundamental objects o\[' linguistic anM- 
ysis are signs modeled by typed fea(,ure strucLures 
and ('.onsCrMned by globM 1)rmciph;s. II I'S(\] <lees 
uoC employ phrase strut'.Cure rules. Instead, very 
general dominance sc\]mmat, a are given. Which ar 
gmncn/,s a lexicM head takes is le×i(:ally sl)e<'iii(~d 
iu its SIII~CAT list. Also adjunction is st)ecili<M 
lexically; t,he ad.iuncC is seen as the semanl,ic head 
which selects t, he kiu<l of signs i(, modifies, (,it(: 
modified sign relnains Lit(; synCa(:t, ic head <)f the 
rcsulCing phrase. \[,ong dis(;ance del>endcncies a.r<'. 
handled in Ill)S(| not; in terms of nlovcment; ImC 
via structure sharing of the values of a SLASH I'ea- 
Cure \])er<:olaCing the "moving" consl,iCuenl,. 
The grammar for German follows the version 
(>f IIPSG giv<;n in (Pollard and Sag, l,<)!)4) rat, her 
si:rit:l.ly, deviaCing only in the following aspects: 
• The Sub<:al, egorization l)rineil)h; is given in a 
binary bra.nching fashion. 
• '|'lie arglllllell|, struct, ure of h'.xical heads is en- 
ri(:hcd. Thus gcncrMizadons ronc<'.l'ning (:ase 
a.ssigntnenC att<l argmnen\[; reduci, ion I)\]ienom - 
ena (:an be <:el>Cured in a principled fashion 
(see Ileinz and Mal, iasek (1994)). 
• Verb second posid<m is handhxl by a mecha- 
nism resembling Cite notion of head movemt:nt, 
of (\] B-Cheery. 
2.3 X2MorF 
X2MorF (TrosC, 1991) is a morphoh)gical (-<tin- 
I)<ment bas<;d on two level morphok)gy (Kosken- 
niemi, 1983). In l;wo-level m<)rphoh)gy mor 
1)hol)honology is treated by means of rules (,hal, 
753 
"\['I\]()N 
SYNSEMILOC 
" geht" 
\[VFORM finit - \]TENSE 
present IEAD \[PERS 3rd 
verb LNUM sg 
2AT \[ F F \[CASE not,,.\] // /cxr/IIF'AD /PERS ,'era / \] 
/ ........ /NUM sg J / 
\/ | L sUBCxT <) J 
/ /"ONT \[~NrmXU\] 
t_ nom_obj \[REI, 
walk\] 
(?ONT LACTOR\[I~ \] 
psoa 
1> 
((cat verb) 
(lex "geht") 
(head ((vform finit) 
(tense present) 
(pets 3rd) 
(num st))) 
(concept ~alk) 
(args 
((actor 
((cat np) 
(head ((case nom) 
(pers 3rd) 
(num sg)))))))) 
Figure 2: Lexical Entry h)r "geM" in HPSG and in FUF 
3.1.1 The Representation of Signs 
The process of recasting the original ITPS(\] 
structures in the FUF tbrmalism can best be de- 
scribed by exalnples. In Fig. 2 the htPS(\] repre- 
sentation of the German verb gem (walks) and it,s 
representation in t?UF is shown, exemplifying the 
following mappings of HPS(\] onto FUF: 
• The subtyl)ing of the IIEAI) iS represented I)y 
the cat feature of FUF. 
• SYNSEM),OCICA'rlHEAD is mapped to head. 
• SYNSEM\]LOCleONTbm, is mapped to concept. 
• Instead of subcategorizing for synsem wdues 
a.s proposed in Pollard and Sag (1994) the 
convention of Pollard and Sag (1987) to sub- 
categorize for signs is adopted. 
• Instead of a list-valued SUBCAT feature the 
feature args is used. The correspondence 
between (syntactic) arguments anti semantic 
roles is established by placing the constituent 
under a feature corresponding to its semantic 
role. Thus list manipulation is avoided and 
the structure corresponds more closely to the 
input specification (given in a language based 
on SPL (Kasper, 1989)). 
• The NONLocal feature is dropped. Slash ex- 
traction is handled differently. 
It should be noted that this entry does not col're- 
spond exactly to the actual representation in the 
generator, it serves simply to illustrate the basic 
ideas underlying the transformation. The actual 
implementation additionally allows for 
• the specification of arguments via external 
macros, accounting for a more principled 
treatment of case assignment, argument re- 
duction and slash extraction; 
• a ditferentiation between lexemes and stems 
to account for a treatment of inflection by the 
morphology component. 
The representation of phrasal signs in IIPSG par- 
allels the one of lexical signs; an additional feature 
DTR, S carries the subconstituents o\[" the phrase. 
One of the daughters is the head of the phrase 
(IIEAD-DTIt), its head features are identical to the 
head features of the phrase (llead Feature Prin- 
ciple). The other daughter may be either a corn- 
map between the lexieal representation of a word 
arm its surface fl)rm. Morphology proper on the 
other hand is viewed as a sirnpte concatenation 
process governed by a regular grammar. 
X2MorF augmertts standard two-lewJ mor- 
phology in two ways. First, it replaces the 
contimmtion class mechanism with a feature- 
based word grammar and lexicon. This is 
an important prerequisite for its use in a 
\[hature-based sentence-level processing system 
(see Trost and Matiasek (1994)). Second, it al- 
lows for interaction between two-level rules and 
word grammar facilitating the formulation of rules 
for non-concatenative morphol.actics like umlaut. 
3 The Integration Task 
Although the main components to be integrated 
fulfill reusability requirements (FU F being a fairly 
general and modular generation engine, the HPSG 
grammar being a declaratively written resource), 
integration of these resources into a unified sys- 
tem couhl only be achieved after suitable adap- 
tation. The morphological component of FUF is 
very restricted. Thus it needed to be replaced by 
X2MorF. The available German word level gram- 
mar of X2MorF was rewritten to conform to the 
feature structure notation employed by FUF. The 
two-hwel part stayed unchanged. More substan- 
|ial changes were required to adapt the HPSG 
granLmar. Not only syntactic adaptations to an- 
other feature formalism were needed, but also the 
olmrational characteristics of' FUF had to be ac- 
counted for. Also some of the phrase structure 
information generalized in the form of principles 
could be "compiled" into phrase structure rules. 
3.1 HPSG in FUF 
First experiments to implement ItPSG in FUF 
rather directly showed inetficient runtime behav- 
ior. Since most grammatical constraints in IIPSG 
are expressed via structure sharing, and FUF 
uses pointers to indicate coreferences, most of the 
processing time was spent in following pointer 
chains through deeply nested feature structures. 
Thus the structures have considerably been fiat- 
tened and some aspects (most notably SUBCAT 
and CONTent) have beet, encoded differently. 
754 
pleiNei'll,, a,u ;tdjurlet,, ;i. marker or a filler (realizing 
l,he shcsh l'e&t,lll'e (if l;|le tt(;;ul d;-:uighl, er). l!\]a('h con- 
sl;ituenl, sl;rHcture is (;onsl_d'~:titl0d by {/.\[1 aSsoci;d.e(l 
sel, or (h)mimmce s(:heniai, a and princillles. 
iII)S(l (lis/,inguishes be/,ween s,lz/Ls/,a'#l, bive (',aJ;e- 
gories (Sllch ;1.8 ilOllil8 or ver|)8) ~md ,\[:\[ul, clio,l, al 
(-a,tegories (e.g., (leterininers). Siliec rliilcdonal 
ea, l,egories COl'l;(:.si)ond (;0 ('\[osed word (;lasses, in 
die I" UF iinl)leHlcnt,~fl, ion (,hese ('at, egori(:8 are (:elf|- 
I)ile(t into phrase sl,rlletllre rules. 
'\['hc s.}Litle apl)roa, ch , i,e. t'a.(:t,oring Slll)('.a,teg:o- 
riz~d;ion hiforl:n~l;hm hire l)hrase sl, rllcl;lil'e rilles, 
is l;~ken wi~Iri auxilia.ry a.nd modal verbs and with 
t)tienoilielia, which t\[t;l,y well be rega.r(led as 1;he 
lllanirosl.a.lJoli el ;.i, flin('.l, iona, l cal,egory, l)u/. whi('h 
are ilOl, ext)resse(/ by lexical it;elns bill; by speci;d 
(;oltst.i|,\[lelil, or(lerillg (e.g., verb se(:olld l\[osit, ion in 
(le('.la..ra.(Jve i\[i;~ili el;roses). 
Th(' i, roatineni; o\[" axljunct, ion in I.he I"U l" iinp\[e-- 
inen/;ai:ion rellects l;he rel)resent~d, ion of lmodi\[iers 
ill the. inl)\[lt; \]glllgll;i.g(':. The llPSG vi(~w of ;l.li ;IAI- 
.i,li-'l; ;\[.s l,he seln;ml;ic head selecl,ing I;ho sign it 
|lied|ties, is changed 1;o the view l,hat adjililCl;s ac.l, 
a8 "OlH;ional" a.rgnl/ielll,S ()\[' t, he synt,;wXh: head. 
3.1.2 Encoding of Principles 
M a,uy c, onsl, ra, it\[t,s exl~resse\[t in 11 PS(\] by Iriea, liS 
of priilcipl(':s (e.g., dOlrliil;'~i|ee s(;\[lell\]~t,;l.) :q.l'e ;i,l- 
ready built into t,h(" phrase sl,r\[i(-l;llre r\[ll(!s (:Olil- 
piled el|l; el I, he origin~J gl'a,i\[lll\];:u'. 'l'here re-- 
in;tin, however, I,he inosi; cet\[{,l'al lll'S(~ princ.i- 
pies eonsl, r;~iliing all phra.s0s an(I ensuring tim 
i)roper ili\['\[)rm~ttioll sharing l)el,wo(`-n rnol;her ~uid 
tie~ul (laughl;er. These are insert, od int, o (,he gra, lll- 
nla.r al; l,|le level (cat phrasal-category). The 
1)ra.ii(;hes dispai;(;hing 1,()parl;i('.uiar phrase l, ypes 
ar(' st)e('ilied l;fl, er in ~/,Ii emi)eddod disjuncl, ion. 
(defparameter *phrasal-princ iples* 
' (; ; ; tIEAD FEATURE PRINCIPLE 
(head {~ head-dtr head}) 
; ; ; SEMANTICS PRINCIPLE: 
(concepL {^ head-dtr concept}) 
(args {~ head-dtr args}) 
(index {^ head-dtr index}) 
;;; SLASH INHERITANCE PRINICIPLE: 
(slash {^ head-dtr slash}))) 
l;'ig\[u'e 3: lIPS(; l'rinciph;s ill li'l\]l '' 
Ilow(wer, o\[\[e irriporl;anl; t)rinciph~ of ll PS(\], the 
Sub(::~A;egorizal, ion I'rineil)le ei'iSlll'illg ~he i)rol)er 
rela.l,ionshi 1) bel;we(m tim a, rgllmenl;8 subcaA,ego- 
rized for a, lld the eonstil,uenl, strllg|.llre el I;he 
phrase 81ill nee(is I,o \])e ;%(;eOlltll, e;d for. Ilow t, his 
const;rn, inl; is /nel, will I)e discussed next. 
3.1.3 Cont;rol S f, ra f, egy 
FUI" el\[lploys a, t, op-down l\[roccssinp; scheme 
(Iriven by l, he synl,;t(:l;ie (;a.I;(:gory ()1: l;lie lnol,her. 
This (-ont;rol sl;ra, t;egy is init(iequal;(" when die con- 
sl,ilJuel\[t st,ruc.I ure is Sl:,e(:itied lexica,lly hy t;he lle;:,A 
a,n\[l l,hus uldcnown u.I;il l, ll(', head is eXl);~il(le(l. 
llPS(\] lends il;self best to heiul-driven, t)otJ;olu-u l) 
i)roc.essiugl a,l, le;-~sl; for genera, l;h)l~l. ,':;ilt(:e t, he COll- 
trol regil\[l(; of b'lJ l" (:anllol, b(' (:ha31ged in prhl(-iph'~ 
(only delay lrle/,hods a, re awdli~ble), l, tie grPdicilri;u' 
it:self has 1o aCCOlltll, for :-ulequa, l,e processil~g ch,u,r- 
acterisl;ics. This i'rie,~l, ns, l,h~t, l,hc h;xicon drive. 
apf)i'oa.ch has I;o 1)e einula/,ed wil,hin t, he g,i'a.Iliiliar, 
based on |;he oper;tl, ionM behavior of FUF. 
The tmsic ido;~ for rcMizmg head driven pro-. 
c.(;ssing behavior is 1,o ilse Idle cset a,nd pattern 
special at, t.rib ul, es of F U F in an asymmetrical fash- 
ion. Ge.eral, ion of a phrase sial'is by realizing 
i{.s head-dtr. 'l'hel'e\['ore only the head daught.er 
is specified ill IJm consl;it, uenl, set, o|' tdlc phra,sc. 
Once l.he \[cxica.l hea.d of l.he phrase is general.ed, 
its ;-trglliiieill, list, is a(-I;iwfl,ed using l, he defaull, re 
(;llrSiOll sl;ra, t, egy o\[" FUF (8inee no cset atl;ril)ut(? 
is l)l'esent). The lexie;dly 8t)\[;(:ilie(I ;trglltnenl.s ;u'e 
uow genera%ed in ~* (virtually) bol;|,onl u l) l'ashion. 
,ql;rll(;t.llre sh;u'ing lmr('oliLi;es l;he args upwa.r(ts 1,() 
the lJiras;tl level, where l;hey m'e l,tieri re;dized via 
Idle lia.l;/,erl\[ \['e;q\]Air(~. The basil; iii(ich&\]:liSll\[ o\[' ell- 
( (cat phrase) 
(head-dtr ((cat lex-cat) ... )) 
;; percolate arguments 
(args {^ head-dtr args}) 
;; :t'ecursion only on head daughter 
(cset (head-dtr)) 
;; realize head and arguments 
(pattern (args head-dtr))) 
Figure d: llead driven genera.lion in FIJI" 
(:oding this processing st, rategy in tim gr;tt\[llila.r is 
given in Fig. 4. If flmcl;ional categories are 1)resent. 
in ~L phrase, /,hen the ~q)l)roprial, e slots have lo he 
81)e('ili(:(l and ;~xhled to cset a, nd pattern. 
Thus Ge shape of I.ll('. resull.ing phrase la.rgely 
depends o, the kind of argutnenl, s it, s lexical head 
adtnit, s. In order to realize its argutnents, every 
word aMe, 1;() acl; as th.:: head of a, phra, se Ira8 1;() pro- 
vide ;t 8ynt;act, i(" a,n(l ,sema\[d;ic Slw~cili('at, ion (:4 i/,s 
a, rgumenl,8. 'this SF.(;(-ili(-;~l;i(m a,180 has l;o a,(;c,:mnl; 
for long distance phenomelm, i.e. (:xl;r~(;I;ion of 
a.ll arg/llllenl; (e.g., wh lnOVel\[lell(,), l,'url,hern\[ore, 
wu'ia, I;ions of case assig.irlent (e.g., in I)assiviza 
l,ion) Inwe 1,o be ;~ccount(xl lot. 
3.i..4 Argmne:n.t; St:rm'~m'c, Encoding 
Although a, large amount ()t' infl\[rmal, ion has I,o 
be 8(x)re(l hi (:tie h:xh;on, a. COlllp~-l(;\[, ;(lid easily 
)uaiitt,~fin~fl)le s/,rucl;ure of (,tie lexicon is a eruchJ 
requh'emenl;. There\[ore ext;ensive Hse has been 
inade of \["UPs external liiacros. 
Fig. 5 stiows l, lic a(;l;\[IM ellcOdil\]g of the lexica, l 
entry for "warlen" ("wait"), 8ubcategorizing for 
all acl, or a.nd a pal, ienl;..qynl,;~ct, ic resl;rict, ious on 
i, lie ;/,i'glllii(;lll, a, re given hy illaoros, pp-allf-acc 
(~Xl)a, nds I;o a 1'1' wii;h l:U'el)oSition a'uf a.lid ILc-- 
('.usa.l,ive case, IJie realizal, ion of die si, r\[let, ural 
~/l'I_~lll\[lellt) llp-ext-da (l('F, en(/s on whel;her i\[rgll- 
755 
(cat lex-verb) 
(ixm "wart") 
(concept wait) 
(args ((actor #(external np-ext-da)) 
<patient #(external pp-auf-acc))))) 
Figure 5: Lexical Entry fl)r "warren" in FUF 
ment reduction (i.e. passivization) has to be per- 
formed or not (for a theoretical background see 
Heinz and Matiasek (1994)). In active contexts it 
becomes the subject and receives nominative case, 
in passive contexts it may be optionally realized 
as a PPvon (see Fig.6). 
(alt ((({^ ^ reduction} no) 
(cat np) 
;; promote to subject 
({" actor} {^ subj})) 
;; passivization 
(({^ " reduction} yes) 
(alt (;; optional pp(von) 
((concept GIVEN) 
(cat pp)(adpos ((lxm "von")))) 
((concept NONE) 
(cat NONE)))))))) 
Figure 6: Expansion of #(external np-ext-da) 
A mechanism common to all arguments and 
thus incorporated into every macro expanding to 
an argument specification is the extraction mech- 
anism required to handle movement (see Fig. 7). 
At the phrasal level the argument which has to 
(;; try to fill slash by unification 
(({^ <slot>} £^ slash})) 
;; does not unify --> add pattern 
(({^ pattern} (... <slot> ...)))))) 
Figure 7: Slash extraction (slightly simplified) 
be extracted (e.g., in wh-questions the constituent 
asked for) has to be specified as tire slash fea- 
ture of the args. Each argument must be checked 
during generation if it is unifiable with the slash 
specification, and, if so, it has to be made coref 
erential with slash. Otherwise, an appropriate 
pattern feature has to be produced to ensure the 
realization of the argument at the args level. 
3.1.5 V2 and a Generation Example 
(\]erman is commonly regarded as an SOV lan- 
guage. However, the standard word order - a sen- 
tence final verbal complex with the finite verb as 
the last dement - is encountered only in subor- 
dinate clauses. In declarative sentences and wh- 
questions the finite element of the verbal complex 
occupies the second position in the sentence. Sen- 
tence initial position of the finite verb is encoun- 
tered in imperative clauses and yes-no questions. 
In our grammar, the verbal complex is always 
generated in the standard order. To account 
for VI and V2 phenomena, a mechanism resem- 
bling the GB notion of head movement is imple- 
((cat s) 
(s-type declarative) 
(head-dtr((cat vk) 
(head ((vform fin))) 
(head-slash ((cat lex-verb))))) 
(v2 {^ head-dtr head-slash}) 
(subj ((head ((case hem) 
(num {^ ^ ^ head hum}) 
(pers {^ ^ ^ head pers}))))i 
(args ((subj {^ subj}))) 
;; force extraction of one constituent 
(alt (((focus GIVEg) 
(focus {^ args slash})) 
((focus {- subj}) 
(subj {^ args slash})))) 
(cset (head-dtr)) 
(pattern (focus v2 args head-dtr))) 
Figure 8: Declarative Main Clause in FUF 
mented. This mechanism flmctions analogously 
to the slash mechanism presented above. If a fea- 
ture head-slash is passed to the verbal temple:, 
the finite verb is extracted, allowing the govern- 
ing phrase to realize it in first or second position. 
The morphology component ensures that separa- 
ble prefixes are left, in place. 
The vert)al complex is generated top down. 'l'he 
arguments of the main verb are generated lexicon 
driven, once the lexical head of the phrase has 
been established. 
Subject-verb agreement and nominative case as- 
signment is handled via the subj slot which is 
coreferential with args: subj and - after argu- 
ment generation -contains the subject of the sen- 
tence (of. Fig. 6). Verb second position (:an only 
be ensltred, if the constituent in sentence initial 
l)osition is nonempty. The slot focus is designed 
to hold that constituent. The constituent to be 
topicalized or, if not specified in the input, the 
subj is extracte<t via the slash mechanism (cf. 
Fig. 7). The interaction between top down cat- 
egory driven and "bottom up" lexicon driven pro- 
cessing is illustrated in Fig. 9, showing also the 
effects of the two slash extraction mechanisms. 
Dcr Be:unto 
focus 
TOP DOWN hat den Brief erhalten 
syntax driven ............. 
v2 i!iiii~ii: vk 
:::::::::::::::::::::::::: • :::::::::::::::::::::::::::::. % head-dtr 
:::::::::::::::::::::::::::::::: • .........,.......,...,..,... 
. ......................... 
::::::::::::::::::::::::::::::::::::: 
::::::::::::::::::::::::::::::::::::::::: e,..,...,.v..,,,,,,.,.. - 
::::::::::::::::::::::::::::::::::::::::::::::: +..,...,....,......,...: i i 
:i:i:i:i:i:i:i:i:::i:i:i:i:i:i:::i:i:::::i:i:i( .... ~_ __' i. ~::~::~::h:~:5:::~:~::~::5~:~:,~ ....... i'h~ad-, lash,  iiiiiiiiiiiiiiiiii:i:i:i:iiiii:i:i:iiiiii:!ili:i:iii:i:iiii!i 
::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::: 
..........,,..,.......,...........,,.,...,.,.,,,.,...,, ====================================================================== 
.-,~.~.~,~.~.~.~.~..,...,...........,..,:,.....,... 
============================================================================== ~:?'"':':::':""'"":':":":::":::"::'"'~a: 
:::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::: • ....................... 
| ........ I "'""""'"'""'"'"''"'"'"'""":"'"' 
Figure 9: Generating a l)edarative Main Clause 
756 
3.2 X2MorF in FUF 
For the integration of X2MorF into I:UF the uni- 
fication engine used in X2MorF was replaced by 
I"UF itself, and the existing word grammar and 
~norph h;xieon were reformulated in the FUF for- 
laalism, and tim word form g('neration l, ask is 
now 1)erfontwd l)y FUI" itself. The= two-level rules 
could be taken over in l, heir original form, only l.he 
morl)hologieal tillers had to l)e l, ranstaLed. 
A simph" functor/argumenL scheme is suflicient 
for the word grammar. The possible eombina- 
l,ions are given I/y the phrase structure rules of l.he 
morl)h grammar. '\['he affixes (fun('t.ors) may fllr- 
the.r restrict the arguments they may lie applied 
t,o. Fig. 10 shows an examph; of morphoh/gical 
e;Ltegorics reslIonsible \[br nominal inlh;ction. A 
u(mn stem has to be \[bllowe.<l by a case suffix which 
determines ease and number ()\[' the resulting noun 
form. 'l'he, head features of the argumen\[, are made 
availal)le t,o the functor via the arghead feature, 
thus <'aiM)ling the funcl, or \[,o subcateg<)rize for its 
argumenl; (e.g., by re(luiriug a certain inflection 
\[>ara.digm). ()he of tim l)ossible case suffixes is a 
troll morl)h inducing l)lural in a certain (;lass of 
nouns with (noun-paradigm null). It applies in 
all cases ex(:el)L dative 3 setting the umlaut feature, 
which U:iggers the two level rule forcing umlaut. 
An example is "(larleu," with plural "G:irlen". 
((cat noun-form) 
(functor ((cat case-suffix) 
(head {^ ~ head}) 
(arghead {~ ^arg head}))) 
(arg ((cat noun-stem) 
(stem {^ ^ stem}))) 
(cset (art functor)) 
(pattern (art functor))) 
((cat case-suffix) 
((lex .... ) 
(head ((umlaut aou-umlaut) 
(case not-dat) 
(num pl))) 
(arghead ((noun-paradigm null))))) 
Figure lO: Nominal \[nlleetion 
'Fhe i\[iterface I>el, we, en synl.act;ic and wor<\[ tew'\] 
t~roeessing is provided hy the \]emma lexicon. It 
<-otd;ains l.he argulnent st;ru<;l;ure of t;he \]exemes 
aim links them I;o (l)ossibly prelixed) stems. The 
require<l syntactic features (>f a particular word 
form arc dct<'a'mined l>y the sentence level syntae- 
ti<'. general,ion. 'Fhe \]elrlll:la lexicon l)aSSeS I;hese 
lhatures to the morl)hoh)gica\] level and l, he woM 
level gr~uttH, art, akes ca.re of sele(:l.ing the al)propri: 
at:e allixes. I)uring the finM linearizati(>li the cx- 
l;ended ewe l<wel rules map l, he <:on<:aCenated stems 
and affixes to the appropriate surface strings. 
:~Thc boolca.n combinal, ions of ce.rtain fea.turcs ll~Lve 
Imcn spelled out in the type hierarchy. 
4 Conclusion 
Wc have shown how existing resources can be 
adapted to new applications therel)y saving eon-- 
siderably on develol>ment efforts. We have demon- 
sl.ral, ed integration tasks on two different levels: 
* integration of software sys/,elns: by c()mt)itJ- 
ing FUF with X2Morl,' we have extended tin2 
flmctionality of FUF. While the original roof 
t)hoh/gy component of FUF ix geared towards 
F, nglish ordy, X2MorF can be used with a 
wide range of languages. 
• Adaptation of linguistic resources to process- 
ing requirements: by adapting our existing 
Ill)S(\] grammar for (~mman to Fur we have 
shown t.hat a de<'laratively writ;ten linguistic 
l'eSO/ll;('(~ can I>e used in a new l>rocessing en 
vironnmnt with modest effort.. 
'l'his is an inq)orLant, st, e l) in bringing natalra\] lan- 
guage processing techniques closer to real-world 
al>plications, where the minimizing of adaptal, ion 
<:<>el; and the maximal use of existing resources is 
crucial for success. 
References 
lhmhbcrgcr, I!',., 1!;. Ga.rncr, W. lleinz, J. Ma.tiasck, 
and 13. l'fahringer. 1991. Vll'\]-l)lJ l)iah>gue by 
Uniti<:ation. \[n 11. Kaindl, ed, 7. Ostcrr. g I 7h.qung, 
pp 42 511, Berlin. Springer. 
Elhadad, M. 1997l. \]"UI": The Universa.l Unilicr User 
Ma.nual, Version 5.0. Tc<:ltnica\] report, I)cpl,.of 
Co.II). So., (?olumbia University. 
\]lcinz, W. a.nd J. Matiasek. 1994. Argum(:nt .qtru<> 
ture a.nd C,a,sc Assignment in (A.'rman. In J. Ncr- 
/mnnc, 1(. Netter, and C. \]%llard, cds, (lermau 
i~* llead-l)rivcn Phrase Structure (:rammar. (\]SI,I 
Publications, Stanford, pp \]99-236. 
\](aspcr, l/,. T. 1989. A tlcxible interface for linking 
applications to l)cnma.n's sentc.nce gem:rater. In 
tb'occcdirt:lS of the DAIU~A Speech and Natural Lan- 
guage Workshop, Philadelphia.. 
Kay, Martin. \]979. Functional Gramma,r. In Pro- 
cccdiugs @" the Fifth Annual Mcctin:\] <4 lhc Bcrkc- 
I<:~\] Li~gt~istics Socicl~\]. l~erkc.lcy Linguisti<:s Society, 
lk~rkeley, (\]A. 
\[(oskenniemi, I(. \] 98',1. Two-Level Model for M,rpho- 
logical AnMysis. In Prec. lJUAI-83, Los Altos, CA. 
Morga,n Kaufmann. 
Pollard, C. and 1. Sag. 11987. \]nformation-tlascd S*jn- 
tax and Semantics, Vol. l: \['hndamcntals. (\]SLI 
l,ccture Notes \]3. CSLI, Stanford, (\]A. 
Polhrd, C. and l. Gag. 11994. llcad-l)rivcn Phra,sc 
,S'tructurc Grammar. Univc.rsil, y of Chicago l>ress, 
( ~hi<:ago. 
'Frost, \[\]ara.ht. 19911. X2MORF: A Mort)hologica\] 
(\]OtllpOtlel|L \]~ts(xl (111 Augmented 'Pwo--l~cvcl Mor- 
phology, ht Prec. lJUAi-,9\[, Sydney. 
Trost, llarMd and 3oha.nncs M~ttia.sek. 1994. Mm'-. 
phology with a. Null-Interface.. \[n lb'oc. (70LlN(\]- 
9~, Kyoto, Japa.n, August 5-9. 
757 
