ࡱ> ]_\c N9jbjbSS %h11?5]XXXXXXX8$ 0Tu)ccc2)4)4)4)4)4)4),s+g-|`)XcEccc`)XX c XX2)l@@XXXXc2) U'XX&) )IP(*Predicate system in French and Thai MT Sunant ANCHALEENUKOON Department of Thai, Chulalongkorn University Bangkok, Thailand. Abstract The predicative system of French and Thai are analyzed with a view to machine translation. This analysis is carried out in terms of logic and semantics, as these two languages are quite different at the syntactic level. We then use the notion of lexical units. The lexical units concerning the predicative system play a very important role in machine translation. They allow to neutralize the differences of the syntactic classes for the semantic class. According to the method of Prof. B.Vauquois, multilevel structures are used in MT systems to represent the analysis of units of translation. A multilevel structures is a tree structure where the nodes bear complex labels containing the necessary information to interpret each level selected. These levels are presented by the hierarchy of their deep structure. The general analysis allows us to observe that the French and Thai languages have the same kinds of concepts: process, state and entity , which constitute the derivative types of the predicates. The result of this analysis shows that the greater part of the predicative system in French and Thai is quite the same. We found that there will be very few difficult points in French-Thai machine translation caused by differences in the predicative systems. On the one hand, there are very few differences concerning semantic restrictions on arguments of predicates. On the other hand, the differences in tenses, modes and word order between French and Thai are not a problem for machine translation, as rules can be described and there is no systematic underspecification. ------------------------------------ Introduction In this paper I provide an analysis of the predicate system in French and Thai MT. The predicate is the term we find in the traditional grammar. In Thai we call Ҥʴ / phaak sadaang / . But the notion of predicate for MT will be carried out in term of logic and semantic. So the principle point of this study are the followings. ---------------------------------------------------- I would like to express my sincere gratitude to Professor Christian BOITET, Assistant Professor Sudaporn Laksaneyanavin and Associate Professor Amara Prasitratasinthu for their suggestions and comments of an earlier version of this paper. I am, however, responsible for any mistakes in this paper. 1. The essential notion concerning the predicate system for MT. We can divide this notion into the notion of predicate, the notion of the lexical unit and the notion of the multilevel structure. 1.1. The notion of predicate. In the period of Plato and Aristotle, they describe subject and predicate in terms of logic. This concept was developed by several European linguists especially Tesniere who founded dependent grammar. It seems that his concept of predicate is suitable to the analysis for Mt. By this, the predicate relates importantly to the number of arguments in a proposition. Furthermore, the predicate here must be the lexical predicate, not the pragmatic. For example: He drinks beer. The lexical predicate is drink supporting he as the argument0, beer as the argument1. While the pragmatic value may be interpreted as a). he absorbs beer. b). he is a beer drinker. 1.2. The notion of lexical unit. As French and Thai are quite different at the syntactic level. We then use the notion of lexical units. The lexical unit concerning the predicate system plays a very important role in MT. They allow neutralizing the difference of the syntactic classes for the semantic class. For example: French Thai marcher (v) Թ (V) march (N) Թ (n) marchant (adj) ---- (adj) The above examples show that there is no word equivalent to marchant (adj) in Thai. However, their lexical unit are just the infinitive form marcher Թ which exist in every language. This lexical unit represent also the same semantic class, that means process . 1.3. The notion of multilevel structure. We will call this M-structures. The M-structures is the method of language analysis for MT by Professor Bernard VAUQUOIS. They are used in MT system to represent the analysis of unit of translation. A M-structures is a tree structure where the nodes bear complex labels containing the necessary information to interpret each level selected. These levels are presented by hierarchy of their deep structure. For example: Vous donnez un nom au fichier.  Sentence donner NP VP NP NP (syntagmatic level subj GOV obj1 obj2 (syntax level Arg0 Arg1 Arg2 (logical level  Agent Patient Ben. (semantic level Vous un nom le fichier Pron. Art. CN. Art. CN. GOV. Des. GOV. Des. GOV . In the syntagmatic level, each node bears the syntagm marks and the morphosyntax class. In the syntax level, each node bears the syntax function. In the logical level, each node shows the logical function. In the semantic level, each node shows the semantic relations. The above M-structures shows that French and Thai are very different in syntax level especially the system of tense, mode, words order, part of speech. By this reason we cannot compare French and Thai in the syntax level. We have to compare them in the logical and semantic level. By this level, the syntax contraint disappear,and the M-structure allows the linguists to calculate every level represented in the only one structure. 2. The criteria of the comparison. The following rules are available to analyze the predicate system in French and Thai: 2.1. The semantic type of predicate. We suppose that, in the natural language, all the lexemes can be divided into entity and predicate. There are two types of predicate, the first is process and state and the second is the elementary predicate and the derivative predicate. 2.2. The number of argument. The number of argument is the method to calculate the predicate. The place of argument is fixed with the number 0-3 or 4. The argument 0 corresponds generally to the subject in active voice. The argument 1 corresponds always to the direct or indirect object. The argument 2 corresponds normally to the second indirect object. The argument 3 corresponds to the third indirect object. 2.3. The semantic restriction on the arguments. The denomination of semantic relation or the semantic trait will be used to explain the semantic restriction of all the lexical system such as : agent, patient, human, animal etc. In the logical level, the semantic restriction on each argument may be different according to the nature of the language. For example in French we use: La patience lui manque. In Thai we say ҢҴʹ In these examples show that argument 0 in French La patience is inanimate but in Thai the argument 0 is human. This kind of structure post the problem to MT. 3. The choice of corpus. The corpus available to MT at the present seem to be the scientific text. So two types of texts are selected a). two technological texts. b). a number of French-Thai entry from the French - Thai dictionary. The technological texts consist of a documentary in software computer dBASE III Plus version English, French and Thai. We call it tri-text The second is hypertext which is in the software program HyperCard Some entries from the French-Thai dictionary are randomed in order to make the point of view of the predicate system larger, not only in the scientific text but also in the everyday language. 4. The result of the study. 1010 predicates were studied. The result can be grouped in two: the predicate with the same number of argument and the predicate with the different number of the argument in French and Thai. 4.1. The predicate with the same number of argument in French and Thai. They are: a) The predicate with one argument such as: apparaitre, exister in the sentence. Le nom et ladresse de chaque client apparaissent. Le format decran PFS : File existe. Apparaitre exister  Arg 0 Arg 0 Le nom de chaque client format decran PFS b) The predicate with two arguments such as : creer , afficher in the sentences : Vous avez cr un fichier. Nouvelle pile affiche une zone de dialogue. Crer afficher  Arg0 Arg1 Arg0 Arg1 Vous fichier Nouvelle pile Zone de dialogue c). The predicate with three arguments such as: ajouter, transmettre in Ajoutez un bouton votre pile Bloc-notes. Unit de base transmet la marche suivre aux objets. Ajouter  Arg 0 Arg 1 Arg 2 Vous bouton pile Bloc-note transmettre  Arg 0 Arg 1 Arg 2 Unite de base marche de suivre objets d). The predicate with four arguments such as: parier in Paul parie 100 francs pour lancourse avec Pierre. Parier  Arg 0 Arg1 Arg3 Arg2 Paul 100 francs course Pierre 4.2. The predicate with the different number of arguments in French and Thai. They are: a). 0 argument in French but 1 argument in Thai. Example: Il pleut . =  pleuvoir Arg0 (0) b). 1 argument in French but 2 arguments in Thai. Example: Jean crit. = ͹¹˹ѧ  ecrire ¹ Arg 0 Arg 0 Arg 1 Jean ˹ѧ c). 2 arguments in French but 3 arguments in Thai. Example: Il tapisse le salon. = һپͧѺᢡ tapisser  Arg0 Arg 1 Arg 0 Arg 1 Arg 2 Il salon ͧѺᢡ The predicates with the different argument in French and Thai are always found in the derivative French predicate. Then the following is the recapitulative table to demonstrate the percentage of the similitude of French and Thai predicate with the same and different argument in dictionary , tri-text and hypertext. Nb Arg-frNb Arg-angNb Arg-thDict fr-th % (731)dBase III+ % (122)Hyper-texte % (157)xAa743xyABab718085xyzABCabc141611xyzw-abcd0--x/xy-a / ab3-- 5. Conclusion Finally we can see that the general analysis allows us to observe that the French and Thai language have the same kinds of concepts : process, state and entity, which constitute the derivative types of the predicates. The result of this analysis shows that the greater part of the predicate system in French and Thai is quite the same. We found that there will be very few difficult points in French and Thai MT caused by differences in the predicate systems. On the one hand, there are very few differences concerning semantics restrictions on arguments of predicates. On the other hand, the differences in tenses, modes and word-order between French and Thai are not a problem for MT, as rules can be described and there is no systematic underspecification. References. Chomsky, N. L analyse formelle des langues naturelles. Paris, Mouton. 1968. Coyaud, M. Linguistique et documentation. Langua et Langages. Larousse 1972. Fillmore, C. The Case for Case, in Universal in Linguistic Theory. Bach and Harms, eds. New York Holt, Rinehart and Winston, 1968. Gsell, R. Actants Predicats et structure du Thai. Relations Predicat-Actant(s) dans des langues de types divers, Actio-documents, Paris, Selaf, 1979. Tesniere, L. Elements de Syntaxe structurale, Paris, Klincksieck, 1959. Zemb, J.M. La fallacieuse equipollence du <> et du <>. Le francais moderne, 46 (1978) no4, 1978. Boitet. Ch. Vers la TAO personnelle. Le Projet LIDIA du GETA. Tribune des Industries de la Langue, No3, Vol. 8-10, juillet-octobre 1990. Boitet. Ch. La TAO a Grenoble en 1990. Actes de ROCLING III. Taipeh, Tsing-HuaUniversite, ed, 1990. Boitet. Ch. TAO et IA : un systeme de traduction automatique. Que doit-il et peut-il comprendre?. Convention IA-91, Paris, Hermes, 1991. Boitet Ch. & N. Nedobejkine. Russian-French at GETA : outline of the method and detailed example. Proc. Of COLING 80, Tokyo, Oct. 1980 Boitet Ch. & N. Nedobejkine. Recent Developments in Russian-French Machine Translation at Grenoble. Linguistics 19, Mouton, 1981. Chappuy. S. Formalisation de la description des niveaux dinterpretation des langues naturelles. Etude menee en bue de Ianalyse et de la generation au moyen de transducteurs. These de3eme cycle, INPG, Grenoble,juillet 1983. Clocksin W.F. & C.S. Mellish. Programming in Prolog. Springer, New York, 1984. Colmerauer A. & H. Kanoui & M. van. Caneghem. Etude et realisation dun systeme Prolog. Universite dAix-Marseille II, 1979. Nitaya Kanchanawan. A Study on Ariane System (inthailanguage). Faculty of Humanities, Ramkhamhaeng University, 1982. Nedobejkine N. & B. Vauquois. Etude de la validite du formalisme choisi pour representer la structure linguistique interface. In Bernard Vauquois et la TAO : vingt-cinq ans de Traduction Automatique. GETA & ass. Champollion, Ch. Boitet, ed. Nedobejkine. N. Representation des informations lexicales dans les dictionnaires electroniques. Conference sur la lexicographie calculatoire, 8-11 septembre 1990, Balatonszabadi, Hongrie. Vauquois. B. Description de la structure intermediaire. Communication. Presentee au Colloque de Luxembourg, 17-18 avril 1978. Vauquois. B. Aspets of Automatic Translation in 1979. IBM-Japan, Scientific Programme, 1979. Vauquois. B. Automatic Translation. Proc. Of the Summer School The Computer and the Arabic Language, Ch.9, Rabat, 1983. Vauquois B. & Ch. Boitet. Automated Translation at Grenoble University Computational Linguistics, Vol.11 No.1 (january-March 19850). *********************** PAGE  ' ./QR`eQh{9R!!:#####?$$$$@%j%k%%%%%%&&&&&&'''')(F(G(J() )h+v+?9@9F9G9H9N90J j0JU5 j6U j jU6 >*OJQJOJQJ5CJ M'(>k~N C x fP`?$'x f`<`" ! '['e'((((() ))")#)P)p)q))******* + + +++++++++ +#+&+)+*+.+2+6+9+<+?+@+E+G+L+N+P+R+S+X+Z+a+c+e+g+h+i+w+,o.{../  ^?`aq<`p$$C}"G5RWQ9R" !w!!""":#m########?$Z$$$$$$$$$$$www & Fw$@%k%%%%%%%%%%%%%%&&&& &I&M&N&&&$wwww & Fw&&&&&&&& ' '['e''''''''')(G(]((((( & Fww$ww(((() ))")#)P)p)q))******* + +4 $$l t# #$$$xww & Fww + +++++++++ +#+&+)+*+.+2+6+9+<+?+@+E+G+L+LXL $$l t# #$$$L+N+P+R+S+X+Z+a+c+e+g+h+i+w+,o.{..//20z00Tx$$ $$l t# #$//20z00v11d22p3S44#556I77%88%9&9>9?9J9K9N90v11d22p3S44#556I77%88%9&9>9?9H9I9J9K9L9M9N9hh&`#$x$#0&P . A!"#$%+,D՜.+,h px  'Unknown Organizationem]5: 'Predicate system in French and Thai MT Title'Predicate system in French and Thai MT 6> _PID_GUID'AN{60A0C9 [8`8Normal $CJOJQJmH H`H Heading 1$h<@&5KHOJQJJ`J Heading 2$h<@&56CJOJQJ<A@<Default Paragraph Font,@,Header  9r &)@& Page Number, @,Footer  9r :B`": Body Text $h CJOJQJ4C`24Body Text IndentDR`BDBody Text Indent 2OJQJ.P`R. Body Text 2DS`bDBody Text Indent 3wOJQJ:Q`r: Body Text 3 wOJQJ0o!0BibliohxCJN5h!      5 $.N5!< N9'?!$&( +L+0N9(*+,-./013/N9)2!8$%@T $$ (  NB  S ?  NB @ S ?NB  S ? NB  S ?NB @ S ?NB  S ?NB  S ?  NB  @ S ?NB  @ S ? NB  @ S ? NB   S ? NB   S ?NB @ S ?NB  S ?NB @ S ?NB @ S ?NB  S ?NB  S ?NB  S ?NB  S ?NB @ S ?NB  S ?NB  S ?NB  S ?NB @ S ?NB @ S ?NB  S ?NB  S ?NB @ S ?NB @ S ?NB   S ?NB ! S ?NB "@ S ?"NB #@ S ?!NB $ S ? B S  ?`abcd !!!"""""""##G$H$I$%%%%%N5?t?)0t ? t? t!Y\t\t\tX\txy\tefttIU t tq5t 4qtItI#4t #J4tI# 4t0 t 0 t 0rt u%t G %]t $t$ztdeXtXtd;5|t;e|t;|t   t  t$Et#t"t(.R_DMQW_kw~? C D H I M N P Q X j q r u @!G!H!J!K!Q!T!Y!k!p!y!!!!!!!!!!!!!!!!!!!!!" " """""0"6":"@"""""""""""######.$3$7$F$K$Q$W$\$f$i$x${$$$$$$$$$$$$% %%)%,%1%4%:%=%H%K%P%R%\%_%e%n%&&&&&&&&&&&&&&&&&'''''*'-'2'5'@'D'G'K'U'W'^'`'[*m*&5=5?5O5k} k  # > qxg|'F:Q<QACN P H!J!!!!"0"6"#Z#####K$Q$n${$ %%7%=%''))&5=5?5O5rAA:\Predicate\Predicate\Predicate system in French and Thai MT.docgggAA:\Predicate\Predicate\Predicate system in French and Thai MT.docLawrence Leung!A:\PREDIC~1\PREDIC~1\PREDIC~1.DOCuserAA:\Predicate\Predicate\Predicate system in French and Thai MT.docuser-C:\Predicate system in French and Thai MT.docuser-C:\Predicate system in French and Thai MT.docuser-C:\Predicate system in French and Thai MT.docuserA:\New Sunant.docuserA:\New Sunant.docmlSihir:Sunant.doc (ml rev 1)     *@56>*CJOJQJo(. @56>*CJOJQJo(. @56>*CJOJQJo(1. @56>*CJOJQJo() @56>*CJOJQJo(1. @56>*CJOJQJo(1. @56>*CJOJQJo(. @56>*CJOJQJo(2. @56>*CJOJQJo(2. @56>*CJOJQJo(2. @56>*CJOJQJo() @56>*CJOJQJo() @56>*CJOJQJo(. @56>*CJOJQJo(4. @56>*CJOJQJo() @56>*CJOJQJo() @56>*CJOJQJo()  l/ix/i@ OJQJo( @05050@i05)5N5p@GTimes New Roman5Symbol3 Arial3Times;WingdingsG CordiaUPCTimes;HelveticaS&Cordia New ThaiTimes"1h 'fk*FZZDC+]0d5&Predicate system in French and Thai MTrmlNB  Oh+'0 ( D P \ ht|''Predicate system in French and Thai MT redredededNormaltmlm68mMicrosoft Word 8.0 @R\ @f@Ϥ@@-++x|3x-|#x9 A 8-A<; ,;A88x|+x8*HL5A,@8(;,@8`8!P|a} N A {A |8`8!P|a} N |!!|{x|;x|3x|+x|#x;HK1A|yxHK=AHKMA8cx<8@8Ke|x5@a@x888kBN |!|#x8|+x88880||x0"HIрA|`5A8!P|N !8( @8!P|N 8iHIA|~x(@8!P|N 80xx0"88HI}A|`5@a88cHIMA|fx(@$xHIA8!P|N a8x,|3xAi|cu@,a8}(Kx8ca89((Ah|cuA|;x}$Kx|hx|#xK58}<|}`9)8 888 8,@,AxHH=A8!P|N K8!P|N |}&Ԑ!|#x}Cx(9@X)|;x|3x|+x||x9 :A 8`z*A 8);A>8`?a88?|9!;9 B80x0-8<88HGA|}x5A$x8!|} N Ax8W:HEA.HBiA||xHBuAHBAxHD)A?, TA480};Kx}#Kx0iHGAcxHGAcxHBuA;`8HF}AxHGyAxHC݀AHD݀Ac.8a8HAQA8a8HBAxHAɀA8!P|aN |}&!@|4T>x88K|~x,@8T9AxHAA<x8TK,8>@$T%, A}#KxHCAH8|9$9 B-@8T1OA@a88<8K|~x,AATc>Kx8!|} N p8aXHCA8!|} N |}&Ԑ!`|#x}CxT>T>|+x|wxH@ A|vxH@AH@)ACx888A||x,@8(@ ;H<0;W>8|KHBA|{x(@ ;HTcx: HAeA88< ,T8(A 8`_8<<.8a8HCAxa<88c0H@MAcxx%xxK||x,A HCAcxH@AH)@8BAxHAAAH8|9$9 BAȨ~x;`KuxX8`a\8\800iH?!A,A$80x0i8X8HBAH 80x0i8X8HBA8dx808h0t8dH>A,@! ՜.+,D՜.+,h px  'Unknown Organizationem]5: 'Predicate system in French and Thai MT Title'Predicate system in French and Thai MT 6> _PID_GUID'AN{60A0C980-6125-11D2-87C6-FA888408170E}x8!`|N (@ 8`N |a!|3x|#x||x;;8x|+xxH;A,@<;8||(@;;@8`8!P|aN (A8`8!P|aN |!!||x8`}Cx|;x|3x|+x|#xH;YA|x(@8!`|!N |8D8?n|*.|+.4|(.|).@8J9$5};@.}'A.@(Ax8rKH 8`r~8c~8!`|!N |}&AȐ!a(}YSx}8Kx}Cx|;x|3x|#x|zx),; ;;@@$8`8!|A} N :*:*;*Z}, 9(U:}?b@X| @A 9=| P@@x88KiH@I, 0@I9`rv=knd| X@@x88K9Hx88K),@|:@AT8x|+x~xH91A,@,x~ijx~x8H7A,A48H8H8xxxgx(xKS;,@8`8!|A} N |ܐ!9`a<|+x|#x8<8}]Sx}T>|+xH0рA|yxH0݀AH0AxCx888<8@H/QA|x,@Ha8dxxxxH/AA|x,A a88H/A8a8H1AA#xH0uAx8!`|!N |}&a! |#x|;x)T>T>|~xADxx888H.A,@ xxfxxx8K@px888H.9A|~x,@HaxexxxH.AA|~x,A a8H.A8aH0AAx8!|a} N |!0|x;H0A||x8`H0AH0yA|~xxH0AH5)A|@AL888888axH3Aax88H0ՀAH8`8!Ѐ|N |a!`|#x8tn|;xT>T>|{x888<8`?@ABCDEFGHIJKMNOPQRSUVWXYZ[^Root EntryhQ3xX~E3 FRf `?H1TablehU$3УV3W3Q3ЫXR30GH?E 3 FH3$U`3W5-3WordDocumentGE3جX~X<{|%hSummaryInformationXx( V3E3xF3Q3XL3DocumentSummaryInformationG8XQP3ET3CompObj<E3T0*Q3XX3~GE 3F 3W3U$3S0*Q3(XRX3~{GY3pE30F3`E3hF3E3F3 E3 FMicrosoft Word DocumentNB6WWord.Document.88x<Exhxx8a9A@K-@,A$,A8cx<;Kрa8(A H%Av,A;x8!|N |}&A!`|}x9@A<(AD9AH|;x8|3x@|+x|#x;9 @$8`8!|A} N (A ;8?H8?|9!K9 B80x0-8L8HH% A|c5A 8!|A} N 8a8LK1x8@8DH"UA,;ZA8;F|