A Constraint-based Case Frame Lexicon 
Kemal Oflazer and Okan Yflmaz 
Department of Computer Engineering and Information Science 
Bilkent University 
Bilkent, Ankara 06533, Turkey 
{ko, okan}@cs, bilkent, edu. tr 
Abstract 
We present a constraint-based case flame 
lexicon arctfitecture fbr bi-directionM 
mapping between a syntactic case Dame 
and a semantic Dame. The lexicon uses 
a semantic sense as the basic unit and 
employs a multi-tiered constraint struc- 
ture for the resolution of syntactic in- 
formation into the appropriate senses 
and/or idiomatic usage. VMency chang- 
ing transfbrmations such as morptnolog- 
ieally marked passivized or causativized 
forms are handled via le:xical rules that 
manipulate case Dames templates. The 
system has been implemented in a typed- 
feature system and applied to Turkish. 
1 Introduction 
-Recent adwmces in theoreticM and practical as- 
pects of feature and constraint-based tbrmMisms 
for representing linguistic information have fos- 
tered research on the use of such formMisms in 
the design and implementation of computational 
lexicons (Briscoe el al., 1993). Case frame ap- 
proach has been the representation of choice es- 
pecially for languages with free constituent order, 
explicit case marking of noun phrases and embed- 
ded clauses filling nominal syntactic roles. The 
semantics of such syntactic role fillers are usually 
determined by their lexicM, semantic and mor- 
plmsyntactic properties, instead of position in the 
sentence. In this paper, we present an approach 
to building a constraint-based case Dame lexicon 
for use in natural language processing in Turkish. 
A number of observations tha.t we have made on 
Turkish tmve indicated that we have to go beyond 
the traditional transitive and intransitive distinc- 
tion, and utilize a Damework where verb valence 
is considered as the obligatory co-existence of an 
arbitrary subset of possible arguments along with 
the obligatory exclusion of certain others, relative 
to a verb seuse. Additional morphosynt~ctic, lex- 
ical and semantic selectional constrMnts are uti- 
lized to map a given syntactic argument structure 
to a specific verb sense. In recent years, there have 
been several studies on constrmnt-based lexicons. 
iR,ussell el al. (1993) propose an approach to mul- 
tiple default inheritance tbr unification-based lexi- 
con. In another study by Lascarides et el. (1995), 
an ordered approach to default unification is sug- 
gested, de Paiva (1993) tbrmalizes the system 
of well-fornmd typed feature struetures. In this 
study, type hierarchies and relations are mathe- 
matically defined. They also formalize unification 
and generalization operators between tin(; featm:e 
structures, along with defining well-formedness 
notion that we use in our system. 
2 Representing Case Frame 
Information 
In rlhu'kish, (and possibly in many other lan- 
guages) verbs often convey several meanings 
(some totally unrelated) when they are used with 
subjects, objects, oblique objects, adverbiM ad- 
juncts, with certain lexical, morphological, and 
semantic features, and co occurrence restrictions. 
In addition to the usual sense wu:iations due to se- 
lectional restrictions on verbal arguments, in most 
cases, the meaning conveyed by a. case Dante is id- 
iomatic, with subtle constrMnts. For example, the 
Turkisln verb ye (cat), when used with a direct ob- 
ject noun phrase whose head is: 
1. para (money), with no case or possessive 
markings and a lmman subject, means to ac- 
cept bribe, 
2. pare (money), wittn a non-human subject, 
means to cost a lol, 
3. para (or any other NP whose head is onto- 
logically IS-A money, e.g., dolar, mark, etc.) 
with obligatory accusative markilig ~md op- 
tional possessive meriting, means to spend 
*~IO~tCy~ 
4. kafa (head) with obligatory accusative mark- 
ing and no possessive marking, means to get 
mentally deranged, 
5. hak (right) with optionM accusative and pos- 
sessive markings, mearls to be unfair, 
854 
6. I~(t.s( (head, of'. d) (or a.ny N\] ) whose het~(/ 
is ontologically IS-A hnma.n) with optional 
accusative mill optio,al i)ossessivc marking 
(obligatory only with ba,~), lilea31s lo waste 
or demote, a person. 
On the other ha.rill: 
I. if' a.n ablative ease-marked oblique object de- 
uoting ;m edible entity ix present, then there 
should not be any direct el)jeer, a.,d tile verb 
,,w:m,s to eat a 1,ieec oJ" (the edible (oblique) oD\[\]eet)l 
or 
2. if the abla.tive case-ll,au'ked oblique o|)ject 
does not denote something edible, but rather 
a. container, then the sense maps to to eat oul 
of, with the optio'naldirect (cdil)le) ol)ject de- 
noting the ol).iect eat(,,. 
Ctea.rly such 11sage ha.s impas:t on tln'Jnatic roh: ns- 
sigmn(;nts to va.rious role fillers, a.nd even (m l.he 
syntactic Imhavior of the vcrl) in question (lh'iscoe 
and (;a.rroll, 1994). I"ol: iustance, for the third 
a.nd l'om:th ca.ses ~d)ove where the ob,i('~ct has to 
b(; ol)lig;d, orily case-marked ~c('usa.tive, a. lmssive 
form wouh| not be grammatical fi)r the sense con- 
vey(',d, ?all;hough syntactic;ally yc (eat)is {t transi- 
tive verl). 
Soruetimes verbs require diilhre, nt co,nbin~ttions 
Of o, rglll\[lellts~ or explicitly require tha.t certa.iu ar- 
gumonts ,,ot I)e present. For insta.nce, the verb qa~ 
requires ditDrent kinds of argunnmts del)endi,g (m 
the sense, obligatorily exel uding other argmnents: 
I. a.n ablative casc-ma.rked oblique objcel and 
with no other object, in tim ca s(', f,'~mw. ~aq 
llle;I.ns to devialc J'rom, 
2. a, dative case-marked oblique object and with 
uo other object, ~a~ menns to be s'a'lT~riscd at, 
3. ~m accusative casc-mar'ked direct object with 
no other objecl, qaq llleg-ms lo be cocO:used 
aboul. 
As ~ |iual examl)\]C, when the verh lul 
(ca.tch/hohl) is used with a.. obligalor:q 3 ~'d per- 
son singular agreemenl ;rod active voice, and the 
subjeel is a Otominalized) ,5' with a verb form or fu- 
ture parliciplc, then the sense conveyed by the top 
level ca.se frame is to \]?el like doinq the predication 
indicated by the subject S's case \['r~mte, with the 
agent being tile subject of tiffs embedded chmse. 
As illustrated in these examples, verb sense id- 
io.m~tic usage resolution h~ts to be (lea, It with in a. 
principled way and not by pnttcrn nmtching (e.g., 
~s in 'l'schichold (1995)), when the l~mg, u~tge has 
a free word order, where l)~l, tern matching al)- 
pronchcs could 5dl. In this p~q)er, we present a 
unification-based apl)ro~wh to ~ constraint-be.seal 
case fra, tne lexicon, in which one single mechanism 
dee.Ix with both l)roblents mtil'ormly, q'hc ess(.nti~d 
function of our lexicon is to m~q) bidirectionally 
I)etween a, case frame containing information that 
is sy~fl;acti<', and ~ sem~mtic Dame wifich c~pl, ures 
the predication denoted by the case fr~mw along 
with information ~d:)out who fills what thematic 
role in that predication. 
3 The Lexicon Architecture 
In this section we present nn overview of stru(': 
ture of lexicon entries m~d the nature of the con- 
sire.Juts. 'Fhc basic unit in the lexi(:on is a sense 
which is the inforlm~tion denoting some indivisible 
predication along with the thematic roles involved. 
We generate the (:axe frame of each sense hy uni- 
\['yiug a set of co-oeeurrelme, morphological, syn- 
t~tctic, semantic, ~md lexieal constraints on vert)s, 
their ~trguments. '\]'he lexicon is implemented in 
TFS (Kuhn, 1993) by the disjunction of the senses 
defined by unifying wf-case-frame (well-formed 
ca.so frnme) with each sense: 
wf-case-fra~m < case-flame. 
wf-case-:frame g SENSE#:\[. 
wf-case-frame & SERSE#2. 
w£-case-frame g SENSE#n. 
3.1 Lexicon Entries 
Ea.ch verl) sense ('ntry in l;|le lexicon has the struc- 
ture shown by the feature structm:e matrix in Fig- 
11170 \]. 
V I,H{\]}: 
AI(,GS: 
,qEM: 
\[ CAT: V :'.I'E M: :(.~: hal-. oot 
.,,.,.o.J: \[1 
Figure 1: Structure of ~ c{tse fi'~une lexicon entry. 
The l:e~ture structure for erich synt~Lctic argu- 
ment contains informal*ion about tim morpholog- 
ical and synthetic structure of the syntactic con- 
stituent such as p~trt-ofspeech, a.greemont, case, 
possessive markers, and additional morphological 
m;trkings such an verb form, (e.g., infinitive, par- 
tieiple, etc.), voice (e.g., active, passive, causatiw', 
reflexive, tee|proeM, etc.) for embed ded S's, ;done 
with their own case frames. This structure is sim- 
ib~r to the structure proposed in Laser, rides cl al. 
(:1995). Ilowevcr, instead of classifying argument 
structures as simply tr~nsit, ive, intransitive, etc., 
we need to consider all relewmt elements of the 
l)ower set of t)ossible arguments. For Tm:kish, the 
syntactic constituents that we have chosen to in- 
855 
SENSE-I!ATI SENSE Bg-UNI;A\]R ,~ENSI!-TO tII:I:EI'T-BRIIIE LW X /L 
• I \[ MAN IS El 11 Ill 
(llll,I()\[ll ISIABI, NO hlll,I()\[lI,.()l\[I NO.IIAT OIIl,.Olll 
VEllII-IS-YI~ 
l SENSES 
AND 
SEMANTICS 
I,EXICAI, 
CONSTRAINTS 
l SEMANTIC 
CONSTRAINTS 
CO-()CCUII.I!NCI~ 
MORI',~I(~L()(IICAI, 
AND 
SYNTACI'IC 
CONSTRAINTS 
Figure 2: The portion of the constraint structure ibr a portion of the the Turkish verb "ye". 
elude in the argument slot (for a verb in active 
voice) are the following: 
• subject (nominative NP), 1 
• direct object (nominative or accusative ease- 
marked ~IP), 
• oblique objects (ablative, dative, locative 
case-marked NP), 
• beneficiary object (dative case-marked ~lP, or 
pP with a certain PFORN), 
• instrument object (instrumental case-marked 
gP or PP with a certain PFORIt), 
• value object (dative case-marked NP or PP 
with a certain PFORH). 
In general, there may be more than one instan- 
tiation of the SEM frame for a given instantiated 
set of case frame arguments (and vice versa). For 
instance, for the ye verb discussed above, the ar- 
gument structure for the third case giving rise to 
the meaning to get mentally deranged may con- 
ceivably give rise to a literal meaning in a rather 
improbable context (such as eating the head of a 
fish at dinner - much in the spirit of the two inter- 
pretations of the English idiom kick the bucket), 
or the same semantics may be expressed by a dif- 
ferent surface form. 
3.2 Constraint Arehlteeture 
We express constraints on the arguments in the 
case frame of a verb via a 5-tier constraint hierar- 
chy sharing constraints among the specification of 
other constraints and sense definitions, whenever 
possible: 
NP's that have no case-marking in Turkish. 
1. Constraints on verb features that describe 
any relevant constraints on tile morphologi- 
cal features of the verb, such as agreement or 
voice markers. 
2. Constraints on mou)hological features that 
describe any obligatory constraints on the ar- 
guments, such as case-marking, verb form (in 
the case of embedded clauses), etc. 
3. Constraints on argument co-occurrence that 
express obligatory argument co-occurrence 
constraints along with constraints that indi- 
cate when certain arguments should not occur 
in order resolve a sense. 
4. Lexieal constraints that indicate any specific 
constraints on the heads of the arguments in 
order to convey a certain sense, and usually 
constrain the stem of the head noun to be a 
certain lexical form, or one of a small set'of 
lexical forms. 
5. Semantic Constraints that indicate seman- 
tic selectional restriction constraints that 
may resolved using a companion ontologi- 
cal database (again implemented in TFS) in 
which we model the world by defining se- 
mantic categories, such as human, thing, non- 
living object, living object, etc., along the lines 
described by Nagao et al. (1985). 
Figure 2 illustrates the simplified form of the 
constraint-sense mapping of the verb yc (eal). 
a.a Valeney Changing Transtbrmations 
As we have already stated, we encode senses of 
verbs in active voice unless a verb has an idiomatic 
usage with obligatory passive, causative and/or 
856 
reflexive voices. 2 In order to handle these valency 
changing transfor-mations, we dellne lexical rules 
as shown in Figure 3. 
INPUT ~ CASE I,RAME 
Refl~.xivi/~,ql(-iN: \] Rellexive: \[- 
I little I ~I < t, 
Rellexive:~+ \[" ~ 
c,,.,~,.~.,,,~,,,/ IN: ~- ~ c, ~a v e i 
(~illl sil li V13 ,.~ ~ ~ 1 .I~XICON 
P~ssbi*ali~,rl ( IN: "~ ~ Passive - 
Passive: ~1 t 
I,'igure 3: Valency transforma.tions using lexical 
rllles. 
This ligure describes how a given case fi'ame 
with its syntactic constituents is processed by a 
sequence of lexical rules each stripping off a cer- 
tain voice marker and then attempting unification 
wii;h t;he lexicon for any possible sense resohttion. 
The order of lexieal rules in this figure reflects the 
reverse order of voice markers in 'Purkish verbal 
morphology, a So a given case frame m~y have 
to go through three lexical rules until it finds a 
unifying entry in the lexicon. \[h|itications be- 
fore going through all le×ical rule.s are for (possi- 
My idiomatic)senses which explicitly require w~r- 
ious voice \]na,rkings. Two additional constituents 
are a,dded via these lexieal rules. 'l'he AGI't-13B3 
(agentive object), (\[enotes the equiwdent of the 
by-objecl in passiw', sentences. The sub.icct of the 
senl;ences a causative voice marked verb is indi- 
cated by CAUSER in the seInani;ics fi:ame. Our cur- 
rent implementation does not deal with multiple 
cansatiw: w)ice rnarkings (which Turkish allows), 
or with the rather tricky surface case change of 
the object of causation depending on the transitiv- 
ity of Lit(: causativized verb. In the examples and 
sa.mple rules below, a voice marker can take one 
o\[' I;l||'ee wdues: (i) +: indicates the voice marker 
has to be l;aken. (ii) -: indica.tes the voice, marker 
is not |:M(en (iii) nil: indic~Ltes the voice mm:ker 
must not be taken; this is used only it, the sense 
detiuitions in the lexicon m|d cm~ unify with - but 
not with +. 
2 For instance: birine vurn, ak 
someone+l)NF hit+lNF 
to hit someone 
VS. 
someone-Fl)AT hit+PASS+iN F 
birine vuruhnak 
to fall i, love with someone 
aWe t,a.ve not dealt with the reciproca.l/colle.ctive 
voice marker yet. 
Iex-rule 
IN: 
I)IJT: 
STEM: ~\] 
VERB: |(;AI, IS: PASS: 
\[I{.I,'LX: 
AII,GS: \[DHbOILI: \[~\] |AG-N-O\]\] J: nil 
LABI,-OI3,1: ... 
SEM: \[plu,n): LI~.OLBS: q 
V Enl3: |(;AI.IS: I'ASS: 
AHGS: / I)ln-OBJ: 
| AGN-OILh LAItL-OI)J: 
\[II,OLES: 
Figure 4: q'he simplified passivization rule for 
transitive w~'r bs 
l"ignres 4 and 5 show two of tit(.' simpler lexical 
rules. 
3.4 Examples 
In this section wc present n \[hw exmnples that 
show how one c~m describe a given verb sense. 
For the tirst example tile following constraints a.re 
employed: 
l. VERB-IS-YE is it constr~fint corresponding to 
\[vl.:itm I S't'r0:m: "ye'\] 
2. VERB-TAKES-IqO-PASSIVE-NO-REFLEXIVE is 
the verb constr&illt VEI{B: \[/I,'LX: nil 
3. DIR-0BJ-ttAS-N0-POSS is the morphological 
eonstrainl, \[An C~s: I ~m~-OlIJ: Ivoss ....... \] 
4. DIR-OBJ-IS-ACC is the morphological con- 
straint |AlmS: I~)m-ouJ: I c:as~ ....... \] 
5. NO-DATIVE-OBL-OBJ is the argument co 
occurrence constraint \[An.c;s: I,)AT-O|,L: ,,lit 
6. SUBJEGT-IS-ttUMAig is the semantic constr~fint 
7. DIR-OBJ-HEAD-LEX-KAFA is a lexical con- 
straint |Alms: IDm-Ol3a: IIn.mD: ILItX: "k~t~"\] 
8. SEM-GET-MEIgTt~LLY-DERANGED is the feature 
structure for the semantics portion 
M| 
We can then express the constraint for the verb 
sense by unifying (denoted by g~ in 'FFS) all the 
857 
lex-rule 
F 
I IN: 
OUT: 
V E I1,13 : 
ARGS: 
SEM: 
I VFIZB: I 
AII.GS: 
SIBM: 
-STEM: 
CAUS: 
PASS: 
nI,'LX: 
SUB J: \[~\]\] 
DIILOBJ: nil 
A13L-OBJ: .., 
nOLPS: \[CAUStDI): 
"STIgM: 
CAUS: 
PASS: 
I~.FLX: 
"SUB,I: 
I)IFC- OILI : 
ABL-OIlJ: 
\['rOLES: \[TIIEME: 
Figure 5: 'l'he simplified causation rule for intran- 
sitive verbs 
constraints ~.d~ove: 
SENSE-GET-MENTALLY-DERANGED := 
VERB-IS-YE 
VERB-TAKES-N0-PASS IVE-N0-REFLEXIVE 
DIR-OBJ-HAS-NO-POSS ~ DIR-0BJ-IS-ACC & 
NO-DATIVE-0BL-OBJ g DIR-0BJ-LEX-KAFA & 
SUBJECT-IS-HUMAN g 
SEM-GET-MENTALLY-DERANGED. 
The resulting constraint when unified with par- 
tially specified case frarne entry -an entl:y where 
only tile argument and verb entries have been 
specified, will supply the unspecified SEN compo- 
nent(s). That is, when a partially specified ease 
\[rame such as 
VEIIB: 
AP(;S: 
S'PEM: "yc." 1 PASS: nil 
CAUS: 
Lm~,i,x: nil J I 
CAT: NIASE: -CAT: STEM: 
SUB J: IIEAD: 
I AGR: 
LPOSS: 
CAT: iP -CAT: 
STEM: DIR-OBJ: IlEAD: CASE: 
kposs: 
N 
"&d&rn" 
noln 
3SG 
llOlle 
"\] "kM&" 
~tcc / 
3SG 
IIOIle .~ 
unifies successfully with the given constraint 
above, the unspecified portion will be properly in- 
stantiated with the experieneer being coindexed 
with the subject in the arguments. 
As a second example, consider tile default sense 
of ye corresponding to cat (somcthi~z.q). '\['he con- 
straints are: 
1. VERB-IS-YE is the verb constraint 
2. VERB-TAKES-N0-REFLEXIVE is the verb c-on- 
straint \[vEa,: I I~\],'Lx: ,,ill 
3. NO-DAT-OBL-OBJ is tile co-occurrence con- 
st,'ain~ \[AR.S: I--A'~-O\]~, .... i~\] 
4. DIR-0Ba-IS(optional-ed±ble) is the dis- 
junctive argument constraint 
(Tiffs is just explanatory, see below for how 
this is implemented in TFS.) 
5. kBL-0BJ-IS(opt±onal-container) is the 
argument constraint, 
\[---{P '- ...... 
6. IgST-OBJ-IS(optional-instrument) is tile 
argument constraint 
7. SEN-EAT1 is the tha~lu:e structure for the se- 
mantics portion 
Ants: |Dm-Ol?3: \[\] 
ABI,-OU3: 
UNS'r: 
"PILED: "to eat" 
SNM: ltOLES: / SOUIt'CE: 
LINST: 
In inost eases, there are arguments that are not 
obligatorily required for resolving a verb sense. 
These, nevertheless, have to be constrained, usu- 
ally on semantic grounds. For instance the di- 
rect object is not obligatory for the basic sense of 
ye, but has to be an edible entity if it, is present. 
We handle these constraints by defining a slightly 
more complex type hierarchy: 
argument = noun-phrase I 
case-frame I 
optional. 
optional = optional-edible I 
opt ional-cont ainer I 
optional-instrument .... 
optional-edible = nil I edible-obj. 
edible-obj ~ noun-phrase & IS-A-EDIBLE. 
where IS-A-EDIBLE is a constraint of the t'orrrt 
\[IIEAD: I SEM: edible\]. The optional ablative and in- 
strumental objects are defined similarly. 4 The 
4Note th~Lt the surface case constraints for these 
are defined in the ha,sic definition of the case fl:ame. 
858 
sense definition then becomes: 
SENSE-EAT1 := 
VERB-IS-YE & VERB-TAKES-NO-REFLEXIVE g 
NO-DATIVE-OBL-OB J 
DIR-0B J-IS (optional-edible) 
ABL-0BL-0BJ (optional-container) 
INST-OBJ-IS(optional-instrument) g SEN-EAT1. 
As a more cornplicated exaInl)le employing nested 
clauses, we presenl, below the case frame for the 
last example in Secl, ion 2, where the verb rut 
(catch) is used with a clausal subjecl; for a very 
specilic idiomatic usage. 
VI'H{ D: 
AIIGS: 
SEM: 
( ',A'\[': V J 
.~;'l'l~ M : "tuC' 
A(;lt: 3S(1 
PAS:;: nil 
CAIJS: nil 
ILI"LX: nil 
LV 1"O It,M: futtne-pal ticit~h~\]| 
-Pl{I,)l): "feel like doing"'\] b;,,,..,.: a\]! 
' 
In this case, the sense resoluliou of the embedded 
case frame is also pe~Jbrmed concurrently with, the 
ease flame resolution of the lop-level frame. 
The last example below illustraLes the handling 
o\[' valency changing (;ransfortmttions where lexical 
I:ulcs hal~dle argument slmllling. 
(\]O CIl k a(\[al n \[,at af, n d an 
(Nild man by 
kar~,ya, geqirildi. 
oppo.sile_side pass+thUS 
+DAT +PRSS+PAST+3SG 
('l'he~ child was passed to the opposil,e side 
I)y I, he mtm,) 
The Olltptl(; \['or this sentence is presenl;ed on the 
right. 
4 Conclusions 
This paper has presented a constraint-based lex- 
icon architecture for representing and resolving 
verb senses an(t idiomatic usage in a case ffa~me 
\[\]:a.mework using constraints on different dimen- 
sions o\[' the inh)rmation avMlable. Economy of 
re,1)reselltation is achieved via sharing of eon~ 
straints across many verb se.nse definitions. The 
system has I)ecn inll)hmt(',nte(l using the TFS sys- 
rein. 
5 Acknowledgments 
This research was in part \[hnded by a NATO Sci- 
• ,1 j enee for Stability Phase Ill Project Giant 1 \[ - 
1 -t i~ I,A NGUA(,I,. 
VEI/lh 
AI/(}S: 
";EM: 
,q TI4M : "get" 
CAUS: 
PASE: 
RFLX: 
CAT: 
SUB,h \[~ II\]'~A D: 
I (\]A'I': DAT-OBJ: \[2~\] II EAD: I 
CAT: 
AGN-OI\]J: ~J IIFAI): 
-i'I~ l,'\]): "to p;tss" 
AGI4NT: \[\] 
1{ OIA!Lq: /GOAI,: \[~ 
LCAUSEII: \[~\]. 
NP \[ 
CAT: N = 
STEM: "gocuk" 
CASE: nora 
AGIL: Sag 
I)O~S: ItOlte 
NP 
, S'\['ICM : "k~Ll ~?' 
(JASI": (I;tt 
A(;\[{ : 3ag | 
LoSS: none J 
NP 
-(\]A'I': N 
STEM: "adam" 
CASE: uom 
AGI{: 3st 
_POSS: none 
References 
lh'is<'oe, I!\].J., A. (?ope~stake, and V. de Paiva 
(eds.). 1993. Inhcrita',,ec I)efaults and lhe LeJ:- 
icon. (;ambridge Univcrsil, y I)ress. 
Ih'iscoe, F,. J. and J. (\]arroll 1994. 7bwards Au- 
tomatic l';ztraction of Arqumcnl Structure from 
Corpora 'lhchnica\] Report, MLTT-006, I{ank 
Xerox I{esear('h Centre, Grenoble. 
Kulm, J. 1993. Encoding .Ill),5'(; (;ramma'rs 
in 77%'. lnstitut f/ir Maschinclle Spracllver- 
arbeitung, Universit'~i.I; SLuttgarl,, (\]ermany, 
March. 
l,ascarides, A., T. Briscoe, N. Asher, and 
A. Copestake. 1995. Order hMepeudent and 
Persisl, ent '\]'yped l)efaull, Unification, TechnicM 
Report, Cambridge University, Computer I,a.b- 
Ol?atory~ Ma.rch. 
Nagao, M., a. Tsujii, and J. Nakamura. 1985. 
The Japanese Government Project for Machine 
Translation. In Computatioual Linguistics, vol- 
ume 1 l. April-September. 
de Paiw~, V. 1993. 'Pypes and Constraints in 
I,KB. In lbiscoe el al. (1993). 
Russell, (\] , A. Ballim, ,I. (Su:roll, and S. Warwie, k- 
Armstrong. 1993. A PracticM Approach to 
Multiple Default Inheritance. for Unitication- 
Based I,exicons. In In Briscoe el al. (:1993). 
'\['schichold, C. 1995. English Multi-word l,exemes 
In A l,exical Database. In Proceedings of th.e 
Lexicon workshop of ES,gLLI'95, Sevenlh Eu- 
ropean ,5'ummer School in Logic Language and 
lnformatio~, Augus(, 
859 
