COLING82,~Ho~ck~(ed.) No~h-Holl~ndPubli~hl~Company 
.OAm~mmiml~2 
The Transfer Phase in an English-Japanese Translation System 
Jun-ichi TSUJII 
Department of Electrical Engineering 
Kyoto University 
Kyoto 
JAPAN 
I INTRODUCTION 
We will discuss in ~his paper 
several topics in a machine translation 
system from English to Japanese which 
was implemented by using ARIANE 78 at 
GETA (Groupe dSEtudes pour la Traduc- 
tion Automatique, Grenoble, France). 
The system basically follows the same 
princilpes as the other GETA~s systems. 
First of all, it uses the English an- 
alysis program in common with the other 
two systems, the systems from English 
to Bahasa, and to Chinese. This means 
that the same interface structure of 
English is used for generating three 
different languages. One of the re- 
search objectives is to verify the 
structure by applying it to Japanese. 
As GETAIs basic ideas were explained in 
(I) and (2) in detail, we will concen- 
trate here on the problems in generat- 
ing Japanese, especially those in 
Transfer Phase. 
Our generation process is, as in 
the other GETASs systems, divided into 
four independent phases, Lexical Trans- 
fer (Transfert Lexical - TL), 
Structural Transfer (Transfert 
structrale - TS), Sturctural Generation 
(Generation Structurale GS), and 
Morphological Generation (Generation 
Morphologlque - GM), which are sub- 
sequently executed in this order. The 
first two phases which we call 
"Transfer Phase" are responsible for 
transferrir~ English oriented struc- 
tures into Japanese oriented ones, on 
which necessary opera,ions will be per- 
formed in the sueoeediag phases (GS and 
@M) to p~erate syntactioally and 
morphologically eorreot Japanese. 
Amom~ others, the problems of 
transferring "Valance Structures" add 
"Tense and Aspect Expressions" are dis- 
cussed in detail in 2 ~ 3, not only 
because they are important problems in 
Transfer Phase, but also because these 
problems show us what kinds of 
"semantic" processings are necessary 
(or unnecessary) for transferring 
linguistic structures of two languages 
belonging to quite different language 
families such as Japanese and English. 
They also give us interesting insig~hts 
into the roles of "semantics" in 
natural language processing in general. 
Some of them are summarized in 4. 
2 PROCESSING OF VALENCES 
2-I BASIC SCHEME 
Same syntactic forms in English 
(direct objects, prepositional phrases 
with specific prepositions, etc.) are 
often expressed differently in syntac- 
tic forms in Japanese. It is obvious 
that there are no one-to-one correspon- 
dences between syntactic functions of 
two languages and therefore, transform- 
ing from one language to another, based 
simply on syntactic functions, is net 
sufficient. 
There are two, essentially differ- 
ent solutions for avoiding this dif- 
ficulty. One solution is to set up 
intermediate "meaning n representations, 
through which surface forms of two 
languages are related. This scheme has 
been recurrently adopted, esp~ci~ly by 
Al-oriented researchers. The oth~ ~ne, 
which we adopted here, is the scheme 
called "lexical unit oriented 
transfer", where many idiosyncratic 
phenomena specific to individual 
lexical units are treated by referring 
to the descriptions in the di- 
ctionaries. In this approach, the 
selection of target s~rfaee foetus is 
perf~ largely dependlng on lexioal 
deaorlpti~ in the Bi-li~ D~- 
ctioeary (W), without refeeri~ to 
M3 
384 J. TSUJII 
universal semantic primitives or rela h 
tions. 
The interface structure adopted by 
GETA is called "multi-level analysis 
tree" which is a kind of annotated 
trees where various kinds of informa- 
tion of several levels such as syntac- 
tic functions (SF), logical relation- 
ships (RL), morpho-syntactic categories 
(K) etc. are attached to each node. 
Such annotation is expressed in the 
form of attribute-value pairs (At GETA, 
"attributes" such as SF, RL etc. are 
called "variables". We follow this con- 
vention in the following.) 
Among the variables used at GETA, 
VL-i(i=1, 2 : Valences) and RL play im- 
portant roles in every stage of trans- 
lation (Analysis, Transfer and 
Generation). The whole process can be 
schematized as follows. 
(Basic Scheme) 
(I) The valences of each source predi- 
cate are described in analysis di- 
ctionary by using VL-i. VL-i indicates 
what kind of surface syntactic form is 
required of the elementwhich fills the 
i-th argument of the predicate. Suppose 
that the verb "reach" has the following 
valences. 
(a) 
(b) 
FOR) 
(c) 
reach to NP (VLI:: TO) 
reach NP for NP (VLI:= N, 
reach for NP (VLI:= FOR) 
VL2:= 
In the AS (Analyse Syntactique), the 
initial string of words is converted 
into an annotated tree strueutre by re- 
ferring to these lexical description 
(See Fig. I). 
I reached ab~k for him. 
1 VCL 
Fig. i Result of Analysis 
(2) The TL replaces the source lexical 
units in the trees with corresponding 
target lexlcal units. The target units, 
especially target predicates, have 
their own valences which show in what 
surface forms the i-th arguments should 
be generated. Because different valence 
strucures such as above (a), (b) and 
(c) often lead to different selections 
of target equivalents, the valence 
information is checked during the 
lexical transfer(See Fig. 2). In some 
cases~ simple source predicates are 
paraphrased by composite target struc- 
tures as in Fig. 3. 
reach 
~--~Lexical Unitl=Tassuru 
|JVLI:=NI 
\[Lexical Unit:=Toru 
"~\[JVLI:=O 
\[JVL2:=NO-TAME-NI 
Fig. 2 Descriptions in BD 
VCL /I, 
Fig. 3 argo GOV argl 
I \ Structure to 
reach pp Structure 
for/ ~ Transfer 
VCL 
argl GOV 
SCL NP Nobasu 
I ar,o/~l v%'ar sl cov 
T oru 
(3) The GS and GM actualize each argu- 
ment in the form specified by JVL-i 
(See Fig. 4). 
VCL . 
JVLI:-O 
J~L2 :-NO-T~-NI 
1 
VCL ~ 
Tonl 
Fig. 4 Actualization Of JVL-i 
THE TRANSFER PHASE IN A TRANSLATION SYSTEM 385 
From the above scheme, though it 
is over-simplified in many "points, we 
can see that the surface forms of the 
two languages governed by predicates 
are almost directly associated with 
each other by the descriptions in the 
BD. 
Furthermore, one can consider that 
valences of a predicate describe sur- 
face usage patterns of the predicate, 
and that the BD associates such usage 
patterns of source predicates with dif- 
ferent target expressions. Because 
GETA s multi-level analysis trees pre- 
serve information of various levels as 
much as possible, we can also use the 
information other than VL-i to enrich 
the specifications of usage patterns. 
For example, the usage pattern of 
"take" 
take the initiative in --ing, 
can be specified by referring to VL-i 
of "take", morpho-syntactic category of 
ARG2 (gerund), the specific lexical 
unit "initiative", etc., and this usage 
pattern as a whole will be associated 
with appropriate Japanese expressions. 
As such, we can transfer naturally 
idiomatic, semi-idiomatic, semi- 
semi-idiomatic --- expressions in the 
source into target ones. This facility 
is extremely important for the language 
pairs like English end Japanese, where 
we can hardly expect one-to-one 
correspondence between lexical units 
and therefore, the selection of ap- 
propriate target units is one of the 
most difficult problems in the whole 
translation process. 
2-2 DISCUSSION 
We adopted "lexical unit oriented 
transfer" or "transfer based on usage 
patterns" instead of using any 
intermediate meaning representations. 
It might be worthwhile mentioning our 
attitude toward the latter approach. 
The meaning representation ap- 
proach seems very attractive, but the 
researchers in this framework have en- 
countered great number of difficulties 
in designing a complete set of semantic 
primitives by which subtle difference 
of meanings of all lexical units can be 
expressed. As Boitet (2) pointed out, 
many systems often use source lexical 
units as primitives in their re- 
presentation schemes, though they use 
certain "universal" sets of primitive 
relationships (Boitet (2) classified 
them as "hybrid" systems). However, 
even in such hybrid systems, to de- 
termine a universal set of primitive 
relationships, deep cases for example, 
is quite problematic. Moreover, we 
doubt whether such relationships are 
really useful for generating target 
sentences. 
We can hardly explain without re- 
ferring to the specific verbs "enter" 
and "go", why we say "John enters the 
auditorium" instead of "John enters 
into auditorium", while we say "John 
goes into the auditorium". As for deep 
semantic ease, "the auditorium" plays 
the same role. The only difference is 
that "enter" incorporates the meaning 
of "into" in its meaning but "go" 
doesnJt. Without semantic decomposi- 
tlons of verb*s meanings, we cannot es- 
tablish any rules on deep cases without 
referring to specific verbs, which can 
decide whether "into" is necessary or 
not. If the rules refer to specific 
verbs, the names of deep oases are not 
signigicant because the same deep case 
is differently interpreted depending on 
indivldual verbs. Why don t you use 
ARGI, ARG2 etc. instead of AGENT, INST 
etc ? 
The case relationships are not so 
powerful in selecting translation 
equivalents, either. If we don~t use 
semantic primitives only by which ap- 
propriate target equivalents can be se- 
lected, we have to refer to the the 
surrounding contexts where the source 
units appear, in order to choose ap- 
propriate target equivalents. Why 
should we reduce the rich structures 
such as multi-level analysis trees into 
poor ones ? We don't claim that seman- 
tic cases are completely useless, but 
only claim that a single level struc- 
ture based on them is not rich enough 
to select appropriate target equiva- 
lents and that surface level informa- 
tion is also useful to specify usage 
patterns (or "contexts where lexical 
units appear"). 
386 J. TSUJII 
3 PROCESSING OF TENSE AND ASPECT 
3-I BASIC SCHEME 
English and Japanese have, of 
course, their own grammatical devices 
to express tense and aspect. As for as- 
pect, for examle, English has basically 
two surface forms, "Perfective" and 
"Progressive", and on the other hand, 
Japanese has the forms 
"PREDicate+AUXiliaries", where AUX is a 
sequence of auxiliary verbs such as 
"Teiru", "Tsutsuaru", "Kake+Teiru" etc. 
However, we should carefully dis- 
tinguish between these surface forms 
(Grammatical Aspects) and what are re- 
ally expressed by them. In the transfer 
phase, we should select appropriate 
Japanese surface forms to express what 
are really expressed in English. In 
order to do this, we set up an 
intermediate representation level which 
is deeper than surface level. The fol- 
lowing five variables and their values 
are used for this purpose. 
I.EASP : Lexical Aspects of English 
Predicates 
NCOM : Non-Completive Verbs 
NMOM: Non-Momentary Verbs 
PMOM: Momentary Verbs 
NTRAN: Non-Transitory Verbs 
PTRAN: Transitory Verbs 
The above values directly 
correspond to the five different 
classes of English predicates shown in 
Fig. 5. 
-STATIVE +STATIVE 
/ / 
t~co_..~s ~O_MM PMO_.MM m"ed~. P'r~ 
Fig. 5 Classification of Eaglish 
Predicates 
2.ESAS? : English Surface Aspectual 
Expressions 
PERF(Perfective Form) 
PROG(Progressive Form) 
3. JASP : Lexical Aspects of Japanese 
Verbs 
RES: Resultative Verbs 
DUR: Durative Verbs 
INT: Instantaneous Verbs 
STATE: Statlve Verbs 
By combining these four values, we 
can discriminate the five different 
classes of Japanese verbs shown in Fig. 
6. 
+STATE -STATE 
(hear) (fail) (sing) (know) (observe) 
Ar-u Kur-u Kik-u Shuppant su- Okor-u 
(exist) (come) (listen) suru(leave) (occur) 
Fig. 6 Classification of Japanese 
Predicates 
4.JSASP : Japanese Surface Aspectual 
Expressions 
Grammatical aspects in Japanese 
are expressed by auxiliary verbs which 
follow the predicates. The values of 
JSASP are such auxiliaries. These 
values are realized as surface au- 
xiliaries in the GS. in some cases, 
more than one auxiliary are needed to 
express the specified DASP(see below). 
TSUTSUARU, KAKARU, TESHIMAU, 
KOTOGAARU, TEIRU, TEKURU, etc. 
5.DASP: Deep Aspect 
UNCOMP: Uncompletion of Activu 
COMP: Completion of Action 
STATE: Absolute State 
EXP: Experience 
THE TRANSFER PHASE IN A TRANSLATION SYSTEM 387 
TDURI: Temporal Duration, including 
the reference time point 
TDUR2: Temporal Duration until the 
reference time point 
IMF: Immediate Future 
ITR: Iteration of Action 
TRANS: Transition of State 
RES: Resultant State of Action 
The basic scheme for generating 
surface Japanese is as foilows. 
(Basic Scheme) 
(I) ESASP (grammatical aapeet) is de- 
termlned in the AS. 
(2) DASP is determined for the combina- 
tion of ESASP a~d EASP (described in 
the dictionary for each English predi- 
cate - lexical aspect). 
(3) An appropriate Japanese equivalent 
for the English predicate is selected. 
(4) JSASP is determined based on DASP 
and JASP of the selected Japanese pre- 
dicate. 
(5) Appropriate auxiliaries with ad- 
equate inflections are generated in the 
GS and GM. 
The above scheme and the detailed 
correspondence among the values are 
illustrated in Fig. 7, and Fig. 8, re- 
spectively. (Fig. 8 shows only the sub- 
,ortion for "progressive forms"). 
Lex$cal Aspect Grammatical Aspect 
Source Verb -M) Source Surface,Expression 
" (-S, +C, . Prosress£ve 
~Deep Aspect o.:d:: el 
Lexleal Aspec: .~ 
Target Verb 
(I g) 
Gra=~at ieal Aspect 
Fig. 7 Basic Scheme 
3-2 MODIFICATIONS IN THE BASIC SCHEME 
The basic shceme can treat the 
following sentences( Here, we will see 
the examples of English progressive 
forms). 
(EX I) He is opening the door. 
Kare-ga Doa-O Ake(Akeru)-TEIRU. 
(he) (door) (to open) 
(-s, -c) ~-- TDURI~---g-(S) 'X ~-..~ (D, R)~ TEXRU 
\"¢(V, -R) I~ ~(I, R) ~ TSUTSUARU 
--(I, -R) I-- 
(-S, +C, -M)~IMF~---------~- (S)----~ KOTO-NI-NARU 
~. ~ YOU-NI-NARU 
\"~(D, -R)--------------- O 
"-,~(I, R)--~0, KAKARU} 
(I, -R)~0 
~.-(S) • 0 
Pro, .... iv° (-s, ~ 
(+s, +T) ~--TDURI~ (s) ~ x \~- 
(v. ~.._~ o X~X ~ t TSI~SU-ARU 
\\-c~, -R~-__.J o Xx~ ' { TEKI-TEIRU 
--(I, -R) ~ X 
'(+S, -T) ~- X 
ASP EASP DAS___.~P JAS.___PP JSASP 
Fig. 8 Rules for Transferring Aspects 
388 J. TSUJII 
(EX 2) The door is opening. 
Doa-GA Hiraki-TSUTSUARU. 
(door) (to open) 
(EX 3) He is leaving. 
Kare-GA Shuppatsushi-KAKE-TElRU. 
(he) (to leave) 
(Ex 4) He is tapping his foot. 
Kare-ga Ashibumishi-TEIRU. 
(he) (to tap one's foot) 
(EX 5) He is doubting his friends. 
Kare-GA Tomodachi-O Utagat-TEKI-TEIRU. 
(he) (friends) (to doubt) 
In these examples, the same 
grammatical aspect in English pro- 
gressive - is realized in Japanese by 
using different grammatical aspects, 
depending on lexical aspects of both 
English and Japanese predicates. Note 
that the same DASP (TDURI) is expressed 
by different auxiliaries in (EX I) and 
(EX 2), because "to open" of transitive 
and intransitive usages correspond to 
the Japanese verbs "Akeru" and 
"Hiraku", respectively, which have dif- 
ferent lexical aspects (Hiraku + TEIRU 
expresses RES, which means "the door is 
open"). 
Though it seems to work well for 
relatively simple sentences, the scheme 
has been augmented in several points, 
in order to treat more complicated sen- 
tences. We will give Just two examples 
of such sophistications below. 
(I) The basic scheme only gives default 
interpretations of DASP. That is, the 
interpretation given in Fig. 8 is ad- 
opted, only if there is no evidence 
which recommends another interpreta- 
tion. Occurrences of time adverbial 
phrases/clauses, for example, often 
change the interpretation. 
(EX 6) He has broken a box. 
(DASP:= COMP) 
He has broken boxes for two 
hours. 
(DASP:= TDUR2) 
We currently distinguish four dif- 
ferent types of such phrases/clauses 
(frequentatlve, duratlve, momentary and 
non-momeltary), and, before the de- 
termination of DASP, a specially de- 
signed subgrammar is executed to 
classify the time adverbials into these 
types. The augumented scheme reflects 
the properties of such adverbials in 
determining DASP. Another example of 
evidences which shift DASP is the 
occurrence of special adverbs such as 
"ever", "yet", "already" etc. 
(2) English to- and ing- clauses in 
predicate valences are expressed by 
subordinate clauses (SCL) in Japanese, 
and we should select appropriate sur- 
face aspectual forms for the SCLts 
which reflect relative time orderings 
among the events described by SCL's and 
the main clauses. 
(EX 7) I saw him walking in the garden. 
.... Arui-TEIRU .... Mi-TA. 
(to walk) (to see) 
DASP of "he walks" is TDURI, because 
the events "I see" and "he walks" occur 
simultaneously. TDURI for "Aruku(to 
walk)" is expressed by "TEIRU", accord- 
ing to the rules shown in Fig. 8. 
(EX 8) I remembered walking in the 
garden. 
.... Arui-TA --- Oboe-TEIRU. 
(to walk). (to remember) 
DASP of "I walk" is COMP, because it 
procedes in time "I remember". 
(EX 9) I remember to walk in the 
garden. 
.... Aruku-null AUX-- Oboe-TEIRU. 
(to walk) (to remember) 
DASP of "I walk" is UNCOMP, because it 
has not completed yet. 
In order to treat above phenomena, 
valences of predicates taking to- 
and/or ing- clauses as arguments are 
augumented with the specifications of 
DASP of the argument clauses, and based 
on these specifications, the same 
scheme as above selects the grammatical 
aspects of the Japanese SCL. 
3-3 DISCUSSION 
THE TRANSFER PHASE IN A TRANSLATION SYSTEM 389 
We emphasize in 2 the lexical 
oriented nature of Transfer Phase and 
claimed that a universal set of case 
relations is not so useful as often 
claimed in literature. On the contrary, 
we set up a set of "semantic" (or deep) 
markers for processing aspeotual ex- 
pressions. Why ? 
First of all, we should notice 
here that, although both EASP and JASP 
seem to describ~ the properties of the 
real world actions which are denoted by 
the verbs, they are just the 
classifications of verbs based on their 
linguistic behaviours in each language. 
When we say that the Japanese verb 
"shinu"(to die) belongs to the class 
(I, R), we don~t claim that the action 
denoted by "shinu" is a momentary ac- 
tion and always happens in physically 
null time, but we only claim that the 
Japanese verb "shinu" linguistically 
behaves in a certain specific way. This 
becomes much clearer, when we consider 
the verb "hiraku"("to open" - intransi- 
tive use) which also belongs to (I, R). 
While the verb "hiraku" behaves in 
Japanese as an instantaneous verb, the 
corresponding English verb "to open" 
behaves as a non-momentary verb (NMOM). 
(Note also that, though "hiraku" is an 
instantaneous verb, we can express 
"Temporal Duration of Action" (TDURI) 
by using the verb in (EX 2)). As such, 
the classifications given by EASP and 
JASP are essentially language-dependent 
and not universal ones. 
DASP, on the other hand, is some- 
what universal. Within the scheme given 
in 3-I, we could omit this variable by 
directly associating surface expres- 
sions in the BD as we did in valence 
transfer. That is, we could associate 
open(transitlve, ESASP:: PROG) 
open(intransitive, ESASP:: PROG) 
Ieave(ESASP:=PROG) 
etc. 
directly with 
Akeru(JSASP:= TEIRU) 
Hiraku(JSASP:= TSUTSUARU) 
Shuppatsusuru(JSASP:= KAKERU,TEIRU), 
etc. 
respectively. However, this direct as- 
sociation method cannot treat various 
kinds of interactions illustrated in 
3-2 between DASP interpretation and the 
other linguistic expressions. We need a 
certain level of representation through 
which linguistic expressions of various 
parts interact. Without DASP, we cannot 
generalize, for example, the influence 
of time adverbials on aspectual 
interpretations. 
Though transferring aspectual ex- 
pressions seems to be performed without 
referring to individual lexical units, 
there are several cases where we have 
to refer to them. This occurs when the 
verbs in the two languages have slight- 
ly different "meaning". The English 
verb "to drown" can be roughly par- 
aphrased as "to die or kill by immer- 
sion in liquid" and, as we can see, the 
meaning essentially contains the con- 
cept "to die" or "to kill". "To drown" 
behaves linguistically in almost same 
manners as "to die". It belongs to the 
verb class NMOM (completive but 
non-momentary). The progressive ex- 
presses form IMF (immediate future) as 
shown in (EX 3). On the other hand, the 
Japanese translation equivalent 
"oboreru" denotes just the real world 
process of one's struggling in water 
not to drown, and behaves as a durative 
and non-resultative verb. Therefore, 
though the two sentences 
(a) He is drowning 
(b) Kare-GA Obore-TEIRU 
(he) (to drown) 
denotes almost same situations in the 
real world, they describe them from 
different points of view, and DASP of 
(a) and (b) are IMF and TDURI, re- 
spectively. The transfer process is 
illustrated in Fig. 9. This process're - 
~xlcal ~it Crammati~1 Aspect 
I.dro~. lJ~sur,.ce~resslo* I .... ~-~(-S. ~, ~ ~ - Progressive I 
~ (D~P of Source ~presslon) 
Target Lexical Unic 
DASP of (, .............. ) 
I - I 
E~resslon of TargeC 
Fig. 9 Rules referring to 
Lexical Units 
390 J. TSUJH 
fers to the individual lexical units, 
"to drown" and "oboreru", and transfers 
"drown+IMP" into "oboreru+TDUR1- as a 
whole. This shows that, even in the 
process of aspect transfer, we need 
lexical-unit-oriented operations. 
Moreover, though we talked until now as 
if EASP and JASP were specified for 
each lexlcal unit, aspectual properties 
of predicates often change, according 
to their usages. Therefore, they should 
be specified for each usage pattern,• 
and aspect transfer should be 
integrated into valence transfer in 2. 
4. CONCLUSION 
We discussed in this paper mainly 
about the role of semantics in Transfer 
Phase by taking examples from our 
English-Japanese translation system. 
The following points should be made 
clear here. 
(I) We can distlnguishtwo kinds of 
semantics in natural language process- 
ings, that is, "semantics as meaning 
representations" and "semantics as con- 
straints (or preference)", both of 
which have their own analogues in 
linguistics, Logical Formula in MG and 
Selectional Restriction Rules based on 
semantic markers. Our contention in 2 
is only that the former type of seman- 
tics is not so useful as often claimed. 
The latter could or should be included 
as descriptors in multi-level analysis 
trees. This is useful not only for re- 
ducing possible ambiguities in Analysis 
Phase but also for augumentlng the de- 
scriptive power of usage patterns in 
Transfer Phase. 
(2) We discussed about the utilization 
of usage patterns in Transfer Phase. 
However, they should be used also in 
Analysis Phase to reduce ambiguities. 
At present, only co-occurrence restric- 
tions between predicates and specific 
prepositions are expressed by VL-i and 
utilized in Analysis Phase, but usage 
patterns of predicates here are much 
more rich, such as co-occurrence of 
specific lexical units (nouns, adverbs, 
etc.), of phrases with specific seman- 
tic and syntactic properties etc. Be- 
cause these are highly idiosyncratic 
.and dependent on each predicate, how we 
can compromise these idiosyncratic mat- 
ters with general rules in Analysis 
Grammar remains as one of important fu- 
ture problems. 
(3) We intensionally avoided the dis- 
cussions about linguistic properties of 
deep cases. In fact, several grammati- 
cal rules can be founded on deep cases. 
By referring to deep cases, we can for- 
mulate, for example, a rule which de- 
cides whether passive construction is 
possible or not. Deep cases in this 
usage give linguistic classifications 
of relationships among predicates and 
noun phrases, but not those among 
events and objects in the real world. 
Deep eases of this type are, however, 
language-dependent as EASP and JASP 
(Rules of passivization in Japanese and 
English are different, for example), 
and therefore, we cannot use them as 
universal relationships in the 
intermediate representations. Moreover, 
even for linguistic deep cases, we 
think that it might be more practical 
to use the other kind Of markers such 
as markers directly showing the 
possibility of passivization etc. 
(4) Though DASP is claimed to be uni- 
versal, this claim should be verified 
in future by applying it to other 
language pairs. In fact, the values of 
DASP reflect many properties specific 
to English and Japanese. That is, we 
set up the values of DASP only to dis- 
tinguish the aspectual features (of 
real world events) which lead to dif- 
ferent surface aspectual forms of En- 
glish or Japanese. These should be dis- 
tinguished in order to transfer as- 
pectual expressions appropriately. Oth- 
er languages might express explicitly 
in surface forms different aspectual 
features of events from differnt points 
of view. Because of this language- 
dependent property of DASP, we perform 
both the interpretation of ESASP and 
the determination of JSASP in Transfer 
Phase. 

REFERENCES 

B.Vauquois (1976):Proc. of COLING 76, 
127-135 

C.Boltet (1980):Proc. of COLING 80, 
430-436 
