An HPSG Parser Based on Description l,ogies* 
J. Joachim Quantz 
Technische Universit~it Berlin, \]'rojekt KHZVM11, FR 5-d 2, 
Franklinstr. 28/29, D-10587 Berlin, Germany, e-mail: jjq @cs.tu-berlin.de 
Abstract 
Ira this paper I present a parser based on De- 
scription Logics (I)I.) for a German lll'SG-slyle 
fragment. The specilied parser relies mainly 
on the inferential capabilities of the nnderlying 
DL system. Given a preferential default ex- 
tension for DL disamhiguation is achieved by 
choosing the parse containing a qualitatively 
minimal number of exceptions. 
1 Introduction 
Ira tiffs paper I present a parser for IIP'SO based on De- 
scription Logics (1)1.). "l)m main mntiwrtion for specifying 
such it parser relies on considerations concerning the dis- 
ambiguation of NL expressions. In ISchmitz, Quantz 931 
it is shown how different types of ambigtfity can be han- 
dled will) a homogeneous approach based on lhe notion 
of preference rules lJackendoff 83\]. A major reqnirement 
for such a t,nified approach is that information usually rep- 
resented rather differently (e.g. syntactic, semantic, and 
encyclopedic information) is homogeneously represented 
in it tin;fern1 and declarative formalism in order to express 
and evaluate the complex l)references stemming from the 
different kinds of information. 
Description Imgics have been developed ira the 
field of Knowledge Representation (see, for example, 
\[Brachman et al. 91\]). qtrey have already been used for the 
representation of semantic and encyclopedic information, 
e.g. IAllgayer et al. 89, Stock 91, Preuss el al. 921. Due 1(5 
their similarity to typcd feature formalisms \[Carpenler 921, 
syntactic infonnation is in principle also expressible in DI,, 
as already sketched in \[Quanlz 93, Quantz, Sehmilz 941. 
Furlhermore, Preferential Defaull Description Logics 
(H)DL) based on weighted defaults \[Quuntz, Ryan 931 can 
be used to represent the preference rules in a declarative 
and formally well-founded way. 
ht the following I will mainly show how IIl,'SG-slyle 
syntactic information can be represented in 1)I., atrld bow a 
simple parser can be build by using tim inference capabil- 
ities of a DL system. Note that when specifying the parser 
I will kee t) the presentation as simple as possible, thereby 
deliberately ignoring efficiency aspects. I will also refrain 
from model;riga//aspccts of relevant knowledge in DL, i.e. 
there are still pieces of information which are not explic- 
itly encoded in the DL modeling, but tire rather ;nil;licitly 
cnntained in lhe parser (e.g. information related to linear 
precedence and to traces}. 
The main objective of this paper it thus neilher to con- 
tribute to research in efficient parsing tectmology, nor to 
"The project KIT-VM11 is suppo21ed by lhe German leedcral 
Minister of Research and Technology (BMFF) under conh'ac101 
IV 101Q8. 
develop a declarative formalism ira which all aspects rele- 
vant for NI.P can be represented. It is rather to prnvide the 
b'ilsis for an implementation of the exception minimiza- 
lion aplsroach to interpretation proposed in IQuantz 931. 
In Section 6 I will hriefly sketch how the l)L-based parser 
l)resenled in Section 5 can be exleuded to realize this ap- 
proach. 
2 ilasie Ideas 
An important distinction made in DI., but missing in tradi- 
tionaI fealure formafisms, is the one between objects ;rod 
/)7)es. I)I. formulae eilher express that a type 1~ is more 
specific than (or subsumed by) a type t2 (tl :< t2) or that an 
object o is an instance of a type or, using I)I, terminology, 
a conce U (o :: c). 
Applying Ibis schema to the task (5t" NI,P, we can say that 
the objects in this donaain are wnrds or phrases, and that 
the types are syntactic categories. I:urthermore, given 
a phrase el we have addiliomll relations between lhis 
l)hrase and its constitnents 02, 03, ..., nsually expressed 
as "02 is a daughter of ol". In DL this ix modeled as 
'(or, 02):: dtrs', or equivalently as '01 :: dlrs:02'. 'dtrs' 
llltlS acts as a binary predicate or, using I)l. terminology, 
as a role. Nile thai roles can have more Ihan one value 
in contrast I(i fealures, which are funclional. We can thus 
write 'el :: dlrs:02 & dtrs:03'. 
Note further Ih;lt the objects stand for occtJrrences of 
words or phrases, and thal different occtwrenees nf the 
slime word will be represenled by different objects. This 
is represented by writing 'o2 :: 15hen:or ', for example, It) 
express thai 02 is tin ocetlrrenee of Ihe form 'er'. 
This is all rather similar R) standard Ill ,.~(; nil;ilion, and 
the main difference is Ihat in addition to the feature struc- 
tures used in IIIHG, we add ;ill additional level of objects, 
which we see as instances nf the feature strueha'es. Fea- 
Itlre structures Ihus correspond Io types or more precisely 
Io l)I. concepts. In a way, IJle objects ill 1)L are used to 
make the lIPS(\] feature slr(lchu'es\]~cr, v\[slcnt, i.e. 1(5 have 
pointers or names to refer to them. 
The additional level of objects allows a slraighlforward 
description of the parsing task. We start with a nttmber 
of objects, namely words, whose phonological value amd 
position ix known. We want to end up with a single object 
containing till these words as (not necessarily immediate) 
consthuents. Now the immediate dominance schemata 
in ill/ IIPSG loll us how 1(5 eonstHlct phrases from words 
or other phrases. Thus the main operation for building a 
phrase is to create a new object being an instance of an ID 
schema (note thilt it) schemata are feature slrtlctures and 
therefore concepts) and I(5 fill in the required daughters 
by us; W the objects available as building mate,'ial. This 
is achiewxt by choosing the 'functor'-daughter' and filling 
the required argumenls. 
412 
Three p(lints are iinporlani in the following sot)ions: 
1. Obviously, 0bjeels cannot be combined in a ran- 
dom way. In III'SG lhe ID schemata and Ihe lexieal 
onlrios eoniahl information concerning c()mbination 
wiih olher ptlrases. T will model lhis infornialion in 
DL tiiid rise standard I)1, itfferoneos lo cheek consis- 
leilcy of eombinalions. Thus lho l)l, sysiom is used 
Io perform Ihe unifleaiion ilisk underlying Ill>S(} and 
similar {Jnification Grainmars+ 
2. All objecl can only be used )is building nlalorial \[+or a 
phrase if il has ilol yel been used iis building maiorial 
for sortie olher phrase. I,'tlrlherinore, when looking, 
for dlitighlers o\[' li now l)hraso, we hay(:. Io fill lhoso 
daughiers for which a filler is required, but not yet 
specified, l will use tile cpi, vtemic o\]scr(t/or k Ill'()+ 
posed in IDonini el al. 921 Io formalize these inlu- 
)lit)f is aud lhon use slandard Ill, rehieval for chocking 
these eonstrainls. 
3+ lq)r synlaclically ambiguous expressions there 
is IIIoro ihan one possibility io coml/ino the 
words/phrases. Since the objects and especially 
ltiO rohitions between them fire viewed from differ 
eli( porspeclive,,i in lhe alloirlalivc' hlterprelali()ns, \vo 
lieed a nioehanisnt {ii Dl, to reproscnl lhose diffefonl 
views, i will i1sc xiluateU de.s'cri\]~tionx 'o :: c in s' 
in Iho followin,~ lo formalize lhis notion of differ- 
ell( porsl)eClives+ Thert~ {s a l+O(ll~h corfespofldeiice 
boD.voon the silualiOllS iisod Io capture the specific 
in)eli)re(aliens and )he charls eroaled in charl pars.- 
iBg. 
3 The IhidcHylnl; Description i ,oEic 
\])eset'iptlon Logics Val y wit the lernl-buildhig operalor:l 
they contain+ In this set)ion 1 will pfesenl lhe ~yillax 
of file 1)1, which is u;ied hi ttio e×ainples given in Ihe 
next lwo seclions, l)ue to space lhnilali(ms I will li(tl 
specify lhe formal seinanlics for this 1)1. (see, for exalilo 
ph', \[I\[oppe et al. 93, Quanlz, Schlnilz 9<!1 for a model- 
Iheorelio SelliailliCS): 
t - ~ e,r, lv,ti&t2,k(i) 
e - > Ilte(r,c) , aOIlle(l') > lie(f) , exaclly(u,f) 
i';() i i I :: 1'2 
r ' > fuai, doi'ilaiil(C) , rillll,e(c) : till2 , iii\'(r) 
"y " ~" I1 :'Q 12 , I1 : : I2 , el : ."- e2 
,:) :: /Y ill S , o>:lt;ild. Sil(51 ,."gO) 
When specifying file fralgmerlt and the pars;el + in the no×l 
,';(+clion.+ I will use a ilolallllil based Oil Ilie PROI.()G inler- 
face pl.,')vide,:t lly the BACK sysIetn lll.l)pe et al. 931. hi 
BACK a dislinclion is lllade belween lerm il~ti'oduclion,'-; or 
dofinilions, and conslritlnl,.liko rlllCS. A lernl ilaino can Do 
inlroduccd cii}ler ilS/~/'/,'lli{il,t" (1, :< l), i,e. ()lily lleCeSSilfy 
(:ondilions are t~,ivon, or as (Iothlod (l, :: : t), i+e. necessary 
arid sufficient condil{oils fire (,iron. A i'ulc Cl > (?2 iliCatlS 
lhal each objocl being at) instal)co of el {s a\],~;o {ill insl;ili('e 
of (;2. 
"lhe foraulhl 'extend sil(sl,s;0' exprc.sses lhe (act thai 
siltlali(m se is ~lri extensicm of siltiali(in sl. This mean:-; 
Iha\[ 'o :: C in sl' implies 'o :: c in ~;7.' for all objeels tl and 
colicopls e. 
It(1 order to dislitiguish belwcen tcllilly, and quelying 
itfformaiioli I will llSO 'o ~: (; ill s' l:or tolls and 'o '?: ill s' 
for qilOi+\[e~. I ftirlhoHnore aSstlnie lhlit ii toll only suceeds 
if it is consistent with lhe previously enlered informal)on; 
otherwise it ill)Is. When the object used in a query is a 
variable, the syslem will relrieve all known instances of a 
concel)l, i.e. 'Object +?: in s' will reltirrl the objects known 
lo be inslances of 'e' in 's' by backtracking. 
Note |ha| the epislemie operator k will ()lily be used 
in queries. It can thereR)re be straightforwardly integrated 
into exisling I)1. syslems. Since this is also true for situated 
descripliorls, lhe parse," presented in Seelion 5 is largely 
b~lsed Oil stil\[ldilrd inference capabilities of DI, systems. 
4 A Snlall l:ragment 
ll'~ lhis section }\[ will present examples from an In .'S(;-slyle 
fragment for German modeled in DL. Due to space lim- 
ilalions I will nol specify all the information contained 
ill Ihis modeling but only lhe one needed 1o illustrale the 
main characterislies of file formalization and lhe example 
sentence 'Die sch6ne l:rau sieht sic' discussed in Ihe nexl 
seclion. 
The fragment is based on Ihe presenlalion in 
\[PcJlhtrd, Sag871 and ils applicalion to German in 
\[llill191L A main difference between lny I)I. modelillg 
and slandard Ill>S(} modeling is lh;l| \]\[ ;IVOi(l fealure imthes 
which would inlroduce st;i)erfluous ITll+ objects. There is 
thus tit) feature 'head' in my modeling since it would yield 
the im,+oduelion of head objects whose ontological sta- 
tus seems controversial. Consequently, my IIead Fea- 
lure Principle specifies cqtdwdence not for a shigle fealure 
'head', but rather for each head feature separately. 
The fragmenl eotlIilii/s five main categories, llamely 
tie(m, ,'lfl, i,e,'D, de/, and tu//. For ilhlslrallon, the definili(ms 
()i II()llll and ,qt'J afe j;iven l)elow: 
II(l{lll : -- lll~lj:ll & lex: I" 
"p : : maj:u & lex: 
Phrase slrtlcltlre is represented by roles as Ihe I'ollowiIig: 
till.'-: :-< dom{lln(si/~il) t~ rallll~o(sl~l,tl) 
COOli)_ dlts :<5 dll.'; 
COlllp dh'\[ :<5 coinp dim & foal 
)lead dlr :< dh~ & teal 
funcior_dir :< dim & feat 
The fealut'e ' funcior, dl r' will be used by lhe parser it) st)co- 
ify Iho sib, n aclin{ ~, as funelor of a now phrase, lls wihle 
will be idenlical to Iho vahio of 'head. dlr', 'adj.dh", ()t 
'filler dh", dependilit; I)ll Ihe p',lrlietllar/##tnlc'(l/{lle Z,)on/- 
/)la,'sc(: (It)) sehemli ilSOd. Nolo lhtil Iho daughlors which 
;ire modehxl as feilltlres ltzt2 funcliomil, i.e. no phrase can 
have lwo fillcr:~ for 'heild dh". 
(;orfest)ondinl_; {o these dau{,Jiler roles and foal(ires we 
have art, umonl roles :lnd fealuros a.<; 'comp_argl' elc, l 
Ihen dislint!,tiish Ill(; R)llowinF~ types of phrase slrilCttlros: 
corllp.:flttlCttlre :: : sonle01ead dh') (~ 
ftiilclor_dh. :\[ioad dlr & 
ilO(ildj dill & lio(liller_dlr) 
adj +'4htlCllll'O : :: sOiile(adj+dh) & seine(head.dill & 
ftilllClOr dh" :adj dll" l~ 
no(coinp.dli.';) & no(filler.dill 
fillcf.shuclute : : sonw,(fillef dlr) tt~ sonic(head dh\] & 
\[:unclor_dh filler dlr & 
no(comp_dll:q) l~ no(adj_dlr) 
Thus hi a 'colnp_sll'tlCItlrt;' Ihe 'head dlr' ,'lOiS as a fililClOr. 
Nolo lhal il has It) be cxplieilly slalod wholher a eerlain 
\['eahlre is enlply, e.g. 'lie(adj..all()' for 'comp_slrueliiro'. 
Ill, syslcms ilSS(llile all ()pell world aild \[like all descrit)iions 
413 
as being partial, i.e. the fact Ihat there is cutxently no known 
filler for a role at an object does not imply thai there will 
never be one. 
"Ilm fragment contains six ID schemata, namely three 
for noun phrases, one for verb phrases, one for adjtmets, 
and one for topicalization. 
idl := eomp~slructure & 
the(headxltr, np & nfo~rn:comm) & 
some(comp_dlrl) & no(compmrgs) 
id2 := eomp~struelt,re & 
the(headxhr, noun& nfom~:comm) & 
no(comp-dtrs) & some(compmrgl) 
k13 := comp~slmctum & lhe(headxlt~.verb) & 
no(eomp-args) & mc:- 
k14 := adj-slrueture & the(adj_dtr, adjunct) 
id5 := tiller_structure & 
lhe(head.dlr, maj:v & no(compmrgs)) & 
me:+ & the(liller_dlr, lop:+) 
id6 := eomp~stmeture & 
the(head~ltr, noun & nform:pro) & 
no(eompxlt~s) & no(comp_a,gs) 
For tile lexieal entries I will use three moq)ho-symtactie 
features (n form, case, gen) to illustrate agreement between 
nouns, adjectives, and determiners. Agreement concern- 
ins ease and gender between nouns and detenniners is 
modeled by specifying that tbe value of tile feature 'case' 
at a common noun is Iho same as the vahle of Iho featnre 
'ease' at the object filing the feature 'eomp_argl' (which 
is the determiner). 
Below are lexical entries for 'frau' and 'sic': 
noun & nfotrn:comm => exactly(1 ,eomp_args) & 
lhe(comp_m'gl ,det) & 
case=eomp_argl .case & 
gen=compmrgl.gen) 
noun & nfoml:pro => no(comp_args) 
lexeme:frau :=> noun & nform:comm & gen:f 
phon:frau => lexeme:frau 
lexeme:er~sie => notnl & nfol'm:pro 
phon:sie => lexeme:er_sie & gen:f 
Note the bierardfieal nature of the modeling--the sub- 
categorization information is specified for common nouns 
and pronouns in general, and is then inherited by each 
specific common noun and pror~oun. Information shared 
by all forms of a lexeme is specified as a property of Ihe 
lexeme, whereas information specilic to a parlicuhw form 
is specified for Ibis form only. 
Adjectives require non-saturated noun phrases as argu- 
ments and agree with them wrt ease .'rod gender: 
adj => adjunct & case=rood-rag.case & 
gen=mod-mg.gen & 
the(mod_mg,np & some(comp.mgs)) 
lexeme:schoen => adj 
phon:sehoene => lexeme:schoen 
Finally, the lexieal enlry for 'sieht': 
verb -----> lhe(eomp_argl,np & case:nora) 
lexeme:sehen => verb & exaclly(2,eomp.m'gs) 
lhe(eomp_arg2,np & case:aec) 
phon:sieht -----> lexeme:sehen 
Note that for verbs taking more than two arguments we 
need addilional features 'comp_arg3' and 'comp_arg4'. 
In addition to tile information modeled so far we need a 
formalization of the principles underlying the combination 
of signs in flING. Some of these principles hold only for 
ph,'ases and not for signs in general. A ph,'ase is defined 
as follows: 
phrase := some(du.s) 
phrase => lex:- 
lex:- => phrase 
"llle llead Fealure Pri, cit;le is then delined as: 
phrase => maj=headxhr.maj & 
gen=head~ht:gen & 
case=head.dt~;ease 
"l'\]m parsing process presented in the next section is essen- 
tially triggered by signs which can act as functors, namely 
signs wilh unsaturated subeat \[isis, signs wilh slashes, and 
pronouns: 
some(args) => funetor 
some(slash) = > funetor 
notre & nfmm:pro =2> funclor 
5 I)L-Based Parsing 
In Ibis section 1 will present the basic structure of a l)L- 
based parser for tile above fragment. The parser is realized 
by five main predicates. I assume that tile initial informa- 
lion given In the parser consisls of descriptions of the 
words occurring in tile expression to be parsed. Consider 
the ambiguous sentence 
(1) Die sch/3ne Frau sieht sie. 
(2) "llle pretty woman sees her. 
(3) Tlm pretty woman she sees. 
"llm initial DI. representation of this sentence is: 
wt :: phon:die & star1:0 & end:l in st 
w5 :: phon:sle & star1:4 & end:5 in st 
Given Ibis information the parser builds phrases from tile 
live words. This is done by creating new phntses until 
no more combim~tions of signs are possible. "l\]m parsing 
succeeds if the words have been all used up and a single 
phrase results: 
paine_sign(Sit,Sit) :- 
findall(Sign,Sign ?: sign & no(k(inv(dtts))) in Sit,\[_\]), 
pm.'s e_s ign (Sit,FinS it) :- 
new_phrase(Sit,NewSit), 
pai~e_sign (NcwSit,FinSil). 
Note that tile epistemic. ¢'oncepl 'no(k(inv(dtrs)))' is used 
to determine whether a sign is still available for phrase 
buiMing. An object is an instance of this concept if it is 
nol a liller of some 'dtrs' role at any other object. 
Tim basic idea of building a new phrase is to look for 
a sign which can act as a funelor, to choose an ID schema 
in which lifts sign is a functor, and to find the required 
;trguments of tile functor. Finally, the linear precedence 
rules are checked and, if necessary, traces are introduced. 1 
new@hrase(SiI.FinSit) :- 
Sign ?: funelor & no(k(inv(dl~,'s))) in Sit, 
selecLid_schem a (S ign,S it,Piu-ase,NewS it), 
eomplete~lrguments (S i gn,NewS it,Nextgit), 
check J ps_and_continuity (Phrase,Sit,NextSit,FinSit). 
Seleetkm of an II) schema is realized in a rather naive and 
simple way---we just take an ID schema and try to create a 
new phrase its an instance of lit is schema, where the feature 
'ft,nctorxllr' is filled by tile funclor. 
IDue to space limitations I do not specify Ihe predieale 
'checkdps_and_continuily' in this paper 
414 
S (p14, id5) 
NP (p8, idl) 
die (wl) N (p/, id4) j',<, 
sch~ne (w2) N (p6, id2) 
H 
lrau (w3) 
S/NP (p13, idS) 
sieht (w4) S/NP/V (pl 2, id3) 
//(tl 1, NP) NP (p9, idS) I! (tl0, V) 
tl 
sic (w5) 
Figure l: Phrase structure of tile sentence 'Die sdffme Frau sieht sic' ('l\]rm pretty wonmn sees her)+ '\]\[' stands for 
'head_dtr', 'C' for 'comp_dtr', 'A' for 'adj_dlr', and 'F' for 'liller_dtr'. 
select_id_schema(Sign,Sit,Phrase,NewSit) :- 
id_schema(ID), 
extend~sit (Sit,NewS\]l), 
Phrase :: ID & ftmetor_tllr:Sign in Newgil. 
Information about existing II) schem;fla thus has to be 
encoded as facts of the form 'kt_sclmma(idl)', elc. Tile 
predicate 'extend_sit(Sit,NewS\]t)' ix used to tell tile DI+ 
system to create a new situation which is an extension of 
tile current situation. 
Note that no further knowledge al)out tile ,'tctual rood- 
cling of It) schemata is uscxl in tile parser except for tile 
fact that each ID schema has a 'funclor_dtr'. Note furlher 
that the DL tell will fail if tile information known about the 
ftmetor cannot be unilied wilh the information required by 
tile ID schema for the filler of 'funetc, r_dtr'. 
In order to complete tile arguments of the functor, tile 
parser cheeks for each argument feature ArgFeat whclhcr 
an argument is required (somc(ArgFeat)) but not yet spec- 
ified (no(k(ArgFeat))). If so, 'lind_arg' looks for such an 
argument and enters it as filler for ArgFeal. Then tile 
remaining arguments are completed. 
complete.argumenls (Functc, r,S il,FinS i1) :- 
arg_feature (AtgFeat), 
Functor ?: some(ArgFeat) & no(k(ArgFeal)) in Sil, 
!,tind_mg(Functoe; Sit,ArgFeal,NewSil), 
complete_arguments(l:unctonNewSiI,HnSil). 
complete_argumenls (_,S it,Sit). 
Again we need to introduce facts sl)ecifying tile arglnnents 
used in tile fragment, e.g. 'arg_fealu re(comp_arg I ) '. 
If an argument is rc*qt, ired it has lo be filled, therefore 
tile Cut, Thus tile recurs\]on lean\]nares successfully only 
when all required arguments are actually tilled. Nol0 thai 
the only information about argument structure needed t)y 
the parser are facts of tim form 'arg_feature(comp_argl)' 
for all argument features. 
To find an argument the parse," looks for a sign which 
has not yet been used for phrase building and ,asserts it as 
filler for Ihe argument feature. Again, if unification is not 
possible due to conflicting constraints (e.g. agreemenl), 
the DL tell will fail. 
tind_arg (Functor, Sit,A~gFeat,FinS it) :- 
Atg '?: sign & no(k(inv(dm-;))) in Sit, 
extend_sit (S it,FinS\]t), 
Funclor :: ArgFeal:Arg in l:inSit. 
tind_arg(Funclot;Sil,Argl:eat,FinSit) :.. 
new_ ph rase(Sit.NewSi,), 
find_a~g(l"unctcw,Sil.ArI,Feat,lqnSil). 
The second clause is needed Io create a required ,'lrgument 
which has not yet been build Ul). Ill this case 'new. phrase' 
is called It) Creale a new potential ;,rgumellt. 
For the sentence 'Die sch6nc Frau sieht sic' we obtain 
two different parses, since bolh 'die schoene frau' and 
'sic' are ambiguotls between nolni/lalive ;ilia accusative 
ease. "llle reading according t() which 'die schoene frm,' 
is subject is shown in Figure I as a t)hrase slructure tree. 
Some of the eorresporvJing infomullion conlained in Ihe 
I)I. situation represenling this reading is given below: 
wt :: phon:die & case:nora & stinl:O ,e¢ end:l 
ps :: idl & head~ltr:p7 & comp~ltrl:wl & 
slarl:O & end:3 
p9 :: id6 & head_dtr:w5 & sla0.:4 & end:5 
IlO :: trilee & tracing:w4 & comp_argl:tu & 
comp.~rg2:p,~ & star1:5 & end:5 
tit :: trace & Iracing:pt~ & slart:4 & end:,l 
I)12 :: id3 &. head dtr:tl0 & 
comp dh'\] :Itt& comp dtr2:p~) & 
slash:w4 & shlsh:pg & slarl:,'l & end:5 
pl~ :: id5 & head dlr:plz & tiller_dlr:w4 & slash:lh{ & 
sI;lll:3 ~ end:5 
p14 :: id5 & heada:ltr:pRs & tiller_dtr:pa & 
stall:0 & end:5 
In the second purse tit and P9 swap places, i.e. l)9 is the 
'comp.dtrl' of \])12 and Ill is the 'comp.Atr2'. 
The rest,It of the parsing process illustrates tile object- 
centeredness of D\]. representations. The constituents of 
tile ulterance are explicitly modeled and can be used 
for extracting or specifying further information. Thus 
we can choose to introduce at feature 'subject' and 
add tile fact '1)1.1 :: subjecl:ps', or we can retrieve all 
tile salurated noun l)hrases (Phrase 2: n t> & no(args)). 
"It,is object-cenleredness is useful fo," disambigualion, 
for example for a,laphor\[i resolution, as ilhtshated in 
\[Quantz, Schmitz 941. 
415 
6 Interpretation as Exception Minimization 
, L • 
I will now briefly sketch how ihe parser presented in 
the previous section can be extended to perform dis- 
ambiguation by exception minimization as proposed in 
\[Quantz 93\]. In case of ambiguous expressions tile parser 
will return more than one situation. Tile basic idea of inter- 
pretation as exception minimization is to model additional 
preference rules nee, ded for disambignalion as DL defaults, 
and to choose the inteqlretation violating a qualilatively 
minimal set of defaults. 
A Preferential Default Description Logic (PDDL) 
based on weigthed defaults has been developed in 
\[Quantz, Ryan 93\]. A weigthed default /5 has the form 
cl "-*,~ ca, where cl is called the premise of/5 (/51,), c2 the 
conclusions of 8 (~5~) and n the weight of/5 (w(8))--the 
higher the weight, the more relevant tire default. For lhis 
PDI)L a formally well-behaved preferential entailment re- 
lation o ~>2 is presented, which is based on an ordering 
on DL models °\[5~:. The basic idea of this preferential 
semantics is to compute a score far each model by sum- 
ming up the exceptions to the defaults. Models with lower 
score are then preferred because they qualilatively mini- 
mize the exceptions. It is straighlforward to carry the idea 
of scoring and ordering over from models to situation. To 
do so, we compute for each situation s and each default 6 
tile exceptions--those objects for which 'Object ?: 6~, in 
s' sncceeds and 'Object ?: /5~ in s' fails. 
If there are several possible interprelations for an ex- 
pression we choose the interpretation given by lhe situa- 
tion with the lowest score. (Note that there may be Iruely 
ambiguous expressions which yield situations with iden- 
tical scores.) Thus taking the above example, we might 
use a preference for topical ization of subjects to prefer Ihe 
parse shown in Figure 1. This can be achieved by simply 
introducing a defat, lt 
np& top:+ ~5 case:nora 
Obviously, this default is a rather weak one and can be 
overwritten by information stemming from selectional re- 
strictions \[Schmitz, Quantz 93\]. 
In principle, it is possible to use preferences stemming 
from weighted defaults already in tile parsing process--- 
situations whose score is higher than a specilied threshold 
are not processed any further. Thus instead of producing 
all parses in tile first step and ordering them in a second 
step, the parser would Ihen only produce tire preferred 
reading. 
7 Conclusion 
I have presented a DL-bnsed parser for a small Ill,(l-style 
fragment of German. Most aspects of tile grammar and 
the parser have been modeled in a highly declarative way. 
Since the main motivalion for my tlresentation has been in 
show how an III"SG parser can be implemented in principle 
by using the inference functionality of a DE system, I have 
deliberately ignored any efficiency issues. It should be 
obvious, however, thai the parser can be further optimized 
to increase its perfornmnce, for example by integrating 
chart parsing techniques. We are currently testing tile 
performance of alternative implememalions of both the 
parser and the underlying DL system. 
One advantage of using DL as underlying formiflism is 
that in addition to the syntactic information modeled ill 
this paper, semantic and encyclopedic informalion can be 
easily integrated inlo tile presenled framework. Furlher- 
more, Preferential Default Description Logics can be used 
to model preference rules as weighted defaults, thereby 
obtaining intcq)relation as exception minimization. The 
l)arser presented in this llaper thus provides lhe basis for 
an homogeneous and formally well-founded approach to 
disambiguatltm based on Preferential Default Description 
Logics. 
References 
IAllgayerct al. 89\] J. Allgayer, R. Jansen-Winkehl, 
C. Reddig, N. Reithinger, "l~,idirectional Use of Km)wl- 
edge in The Multi-Modal NL Access System XTRA", 
in 1JCAI'6'9, 1492-1497 
\[I3,-achman et al. 911 R. t3rachman, I).L. McGuiness, 
P.F. Patel-Sclmeidcr, L. Alperin Resnick, A. 13orgida, 
"Living with CLASSIC: When and How lo Use a KI.- 
ONE-like Language", in J. Sow) ted.), Principles of Se- 
mantic Networks: l;:vplorations hi the Representatkm 
<(Knowledge, San Male(l: Morgan Kaufmann, 1991, 
401-456 
\[Carpenler 92\] B. Carpenter, 7'he Logic of "l),ped l,'eature 
Structures, Cambridge: Canlb,idge Unive,'sity Press, 
1992 
\[Donini et al. 921 F.M. Donini, M. Lenzerini, D. Nardi, 
A. Schaerf, W. Null, "Adding Epistemic Operators to 
Cnucept Languages", KR-92, 342-353 
\[Ililt1911 W. Hiltl, Ein HPSG-Enlwutf fffr &ts Deutsche 
und sehle hnplementierung,, Diplomarheit l.Jniversiliit 
Koblenz-Landau, 1991 
IHoppe el al. 931 T. Hoppe, C. Kinclermann, JJ. Quantz, 
A. Schmiedel, M. Fischer, I~ACK V5 Tutor#zl,~ Manual, 
KIT Report 100, Techniscllc Universitiit Berlin, 1993 
\[Jaekendoff 831 R. 3ackendoff, Semantics and Cognition, 
Cambridge: MIT Press, 1983 
\[Pollard, Sag 87\[ C. Polhlrd, I.A. Sag, An Informatkm 
Baxed Syntax and Semantics, Vol. I Fundamentals, 
Stanford: CSLI Lecture Noles 13, 1987 
\[Preuss et al. 921 S. PreuB, B. Schmilz, C. llauenschild, 
C. Umbach Anapkora Resohttion in Machine 7)')as- 
#1I\[o11, KIT Report 104, Technlsche Universiliit I3erlin, 
1992 
\[Quantz93\] J3. Quanlz, "lntcrprelalion as Exception 
Minimization", IJCA\[-93, 1310 1315 
IQuanlz, Ryan 931 J3. Quantz, M. l~yan, PmferenlktIDe- 
Jhult Description LoL, ics, KIT-Report 110, Technische 
Univcrsitiit Herlin, 1993 
\[Quanlz, Schmilz941 J.J. Quantz, FL Sctunitz, "Know- 
ledge-Based l) isamlliguat ion for Machine Translation", 
Mhtds and Machhles 4, 39 -.57, \] 994 
\[Schmitz, Qt,antz 93\] 1;. Schmitz, 33. Quanlz, De, lulls 
ht Machine 7kanslation, KIT Report 106, Technische 
Universitiit Berlin, 1993 
\[Stock 91\] O. Stock, "Natu,'al Language and Exploration 
of an Inform)lion Space: tile AI .Fresco lnleraclive Sys- 
lem', in IJCAI'91,972 978 
416 
