File Information
File: 05-lr/acl_arc_1_sum/cleansed_text/xml_by_section/metho/73/c73-2030_metho.xml
Size: 23,938 bytes
Last Modified: 2025-10-06 14:11:12
<?xml version="1.0" standalone="yes"?> <Paper uid="C73-2030"> <Title>MORRIS SALKOFF ON USING SEMANTIC DATA IN AUTOMATIC SYNTACTIC</Title> <Section position="1" start_page="0" end_page="0" type="metho"> <SectionTitle> MORRIS SALKOFF ON USING SEMANTIC DATA IN AUTOMATIC SYNTACTIC ANALYSIS * </SectionTitle> <Paragraph position="0"> O. Introduction.</Paragraph> <Paragraph position="1"> A program for the syntactic analysis of French text has been developed which is used in conjunction with a string grammar of French 1 and a dictionary of the grammatical properties of the words in the sentences analyzed. The program has been written in fORTRAN SO that it can be used on any computer having a rORTaAN compiler. It has been tested on the IBM 360-91 at the Centre d'Etudes Nucl&ires in Saclay, and on the IBM 370-165 at the computing center of the C.N.P,.S. in Orsay and yields satisfactory analyses of sentences in a reasonable computing time.</Paragraph> <Paragraph position="2"> The analyses obtained for the first eight sentences of a text by Jacob and Monod in molecular biology are presented in the figures of the appendix. My object in this paper is to discuss what semantic data have been incorporated in the string grammar in order to produce these analyses and to prevent other false analyses from being obtained. A discussion of certain features of these analyses, as well as of those analyses not obtained, will bring out the nature of the semantic data that I have in mind. The text analyzed is reproduced in Fig. 1, and the analyses in the figures that follow.</Paragraph> <Paragraph position="3"> of ann&s. This analysis is a possible one (akhough not correct in this sentence), for a sentence such as Quelques ann&s sur ce probl~me m'ont convaincu que ... is correct.</Paragraph> <Paragraph position="4"> b) The sequence Les connaissances ... acides nucldiques et protdines is the subject in the third and fourth analyses of JM-1 (figs. 4 and 5). c) The pair of parentheses (), e.g. in line 7 of Fig. 2, represent a zeroed indefinite subject of the verb comprendre: (permettent) ?t quelqu'un de comprendre cela ~ (permettent) ( ) de comprendre cela. The double parentheses in line 13 of Fig. 3 represent an element that has been' reduced to zero in the domain of a conjunction.</Paragraph> <Paragraph position="5"> The following analyses were rejected: d) The analysis in which the sequence (1) depuis quelques ann&s .... acides nucl~iques et prot~ines would have been analyzed as a prepositional phrase consisting of depuis followed by a conjunction of three noun phrases. In effect, a concrete noun, a time noun, or a nominalization is possible after depuis: (2) a) Depuis le coin de la rue, (je vous dis cela) b) Depuis ftuelques ann&s, (je vous dis cela) but two different types cannot be conjoined after depuis: (3) * Depuis le coin de la rue et quelques ann~es, (ie vous dis cela) Since (1) is similar to (3) in this respect, this analysis was rejected. Sentence JM-2 Remarks a) bien is analyzed as a sentence adjunct only when it occurs to the right of a form of ~tre, as in this sentence, or in such sentences as (4) c'est bien lui; rid& ~tait bien a moi, b) The sequence a la suite de should be analyzed as a complex preposition, but the appropriate mechanism to accomplish this has not yet been added to the analyzer.</Paragraph> <Paragraph position="6"> R.ejected analyses.</Paragraph> <Paragraph position="7"> c) The prepositional phrase a N cannot be the right adjunct of a proper noun:</Paragraph> </Section> <Section position="2" start_page="0" end_page="0" type="metho"> <SectionTitle> ON USING SEMANTIC DATA IN AUTOMATIC SYNTACTIC ANALYSIS 399 </SectionTitle> <Paragraph position="0"> (5) *Pierre c) Paris (est mon ami) ; mais: Pierre de Paris (est mon ami) This restriction prevents the program from obtaining the incorrect analysis in which ~i la suite ... is a right adjunct of Crick.~ d) the analyser rejects the analysis in which the sequence (6) qu'avaient propos3 Watson et Crick c) la suite is taken as similar to the sequence (7) qu'avaient proposd Watson et Crick au gouvernement The verb proposer belongs to a subclass V15 that requires a human-like noun in the position N~ of its object N1 ~ N~: (8) proposer quelque chose \[au jury I ~ ce#e f~,~me I a la foulel ...\] *proposer quelque chose c) la table The verb apporter also belongs to V15, and so the analyzer does not obtain the analysis in which the sequence (9) apportd la preuve que ... ~ la suite is taken as similar to apporter un chapeau cl Marie. Sentence JM-3 Remarks a) The sequence (10) le d&ouverte de rARN messager is analyzed as a &quot;compound noun&quot; This term was meant for such groups as (11) lyc& d'Etat ; homme de l'espace ; bateau c~ vapeur When the mechanism for treating complex prepositions like ~ la sulte de has been added to the program, it will be specified in the grammar that the prepositional phrase la suite de N must be analyzed as a sentence adjunct.</Paragraph> </Section> <Section position="3" start_page="0" end_page="0" type="metho"> <SectionTitle> 400 MOmUS SALKOPP </SectionTitle> <Paragraph position="0"> in which the prepositional phrase de N or ,~ N cannot be pronominalized to en or y: (12) a) Je vois un lyde d'Etat :#~. *J'en vois un lyc~e b) Je vois un bateau ~ vapeur C/:~ *J'y vois un bateau But this pronominalization is possible for (10): (13) On ddcrira la ddcouverte de rARN messager-+ On en d&rira la d&ouverte and so should not be analyzed as a compound noun. However, some technical difficulties with repeated adjuncts made it convenient to retain this analysis temporarily.</Paragraph> <Paragraph position="1"> Rejected Analysis.</Paragraph> <Paragraph position="2"> c) The analyzer rejects the analysis in which the object of confirmd is (14) a) Les ... hypotheses ... et les prot3ines-enzymes i.e., the analysis in which the sentence is understood as b) (Les d&ouvertes ...) ont confirm3 les hypotheses ... et (les d&ouvertes ont confirmS) les prot3ines-enzymes. The verb confirmer cannot take a concrete noun as its object: (15) a) *\[Pierre / l'hypoth~se\] a confirm~ les prot~ines except for a few special cases such as b) Pierre a confirmC/ \[sa place (dans ravion) / le rendez-vous/ ...\] Sentence JM-4 l~emarks In the analysis shown (Fig. 9), the relative clause qui caract~risent une cellule is analyzed as a right adjunct of activit~s (cf. line 17.), although it should be an adjunct of the entire subject group, i.e. an adjunct of les propri\[t~s, les structures, les activit~s. There is at present no provision in the grammar for attaching an adjunct to a sequence of conjoined structures.</Paragraph> </Section> <Section position="4" start_page="0" end_page="26" type="metho"> <SectionTitle> ON USING SEMANTIC DATA IN AUTOMATIC SYNTACTIC ANALYSIS 401 </SectionTitle> <Paragraph position="0"> Rejected analysis.</Paragraph> <Paragraph position="1"> The analyzer rejects the analysis in which the object of d~montr~ is (16) que C1 et * N,z; C1 = les propri~t~s, les ... rapport&s h la structure</Paragraph> <Paragraph position="3"> N, = des protdines que ...</Paragraph> <Paragraph position="4"> For one, protdines is not a correct object ofd~montr~: (17) *Pierre a ddmontr~ les prot~ines so that the restriction mentioned above in JM-3 in connection with confirm3 disallows the analysis. However, even if N, happened to be a licit noun object of d3montr3 there is still a question whether the conjunction of a nominalized sentence and a noun phrase yields a grammatical object: (is) ? Pierre a d3montrd que le probl~me est diffcile et l'impossibilit~ de le rt;soudre Such a sentence is difficult or impossible to accept; in the present grammar, it is taken as ungrammatical, and a restriction prevents the conjoining of such dissimilar objects.</Paragraph> <Paragraph position="5"> Sentence JM-5 The prepositional phrase par un segment g3n3tique is taken as a right adjunct of the verb affirmer, instead of being analyzed as an adjunct of the participle d3finie.</Paragraph> <Paragraph position="6"> Sentence JM-6A The two analyses show the alternatives for conjoining the sequence ou d'une lign& cellulaire. In the first analysis (Fig. 11) it is conjoined to d'une cellule, which yields the meaning intended by the author: I'ADN d'une cellule ou (I'ADN) d'une lign& cellulaire. In the second reading, it is attached to la structure: la structure de I'ADN ... ou (la structure) d'une a The asterisk here represents a sentence adjunct.</Paragraph> </Section> <Section position="5" start_page="26" end_page="26" type="metho"> <SectionTitle> 402 MORRIS SALKOFF </SectionTitle> <Paragraph position="0"> lign~e cellulaire. The general problem of the correct conjoining of strings headed by a conjunction has not yet been solved.</Paragraph> <Paragraph position="1"> Sentence JM-6B Because the mechanism for treating idioms has not yet been incorporated in the analyzer, it was not possible to treat en fonction de as a complex preposition. It is therefore analyzed as a prepositional phrase enfonction modified by the right adjunct de signaux ... (cf. note 2). Sentence JM-7A Remarks.</Paragraph> <Paragraph position="2"> a) The sequence la conversion de son syst~me excr~toire is analyzed as a compound noun; remark (a) on sentence JM-3 applies here too. Because of this, the relative clause qui, de semblable ... is incorrectly attached to conversion, instead of being analyzed as the right adjtmct of syst~me.</Paragraph> <Paragraph position="3"> b) The exact status of the sequence de semblable ~ celui d'un poisson is not clear. It could be analyzed as a verb adjunct (that can appear at the beginning of a clause or sentence) for a subclass of verbs like devenir, se transformer, ...; or it can be treated as part of the object for these verbs. I have chosen the second solution, and it appears as the first part of a split object of deviendra in an inverted center string (line 13, Fig. 14). c) The adjective excr~toire, since it is neither definitely masculine nor definitely feminine in form, could modify either conversion or syst~me. The analysis shows it modifying syst~me (line 7), and further analyses in which it would modify conversion are not printed by the analyzer, since such an ambiguity is predictable from the first analysis. The same remark applies to thyroidienne, which can modify either injection or hormone. The printing of these ambiguities (and of many others), which can be predicted from the form of the structures involved, is suppressed by the program.</Paragraph> <Paragraph position="4"> Rejected analyses.</Paragraph> <Paragraph position="5"> d) In fig. 14, analogue is taken as the adjective object of deviendra. However, analogue is also a noun: (19) Cette situation est en effet l'analogue (de telle autre) ON USING SEMANTIC DATA IN AUTOMATIC SYNTACTIC ANALYSIS 403 but it cannot appear as the noun object of ~tre without an article. Only a certain subclass of nouns, called N17, can appear without the article: (20) Pierre est \[patron / ambassadeur / professeur/ ...\] *Pierre est rocher Since analogue, as a noun, does not belong to this subclass N17, the analysis is eliminated.</Paragraph> <Paragraph position="6"> e) The analyzer rejects the analysis in which the object of pro-</Paragraph> <Paragraph position="8"> This is analogous to the use ofprovoquer in sentences like (22) On a provoqud Pierre \[~ la violence / ~ un acte d~sdspdr~/ ...\] But provoquer belongs to a subclass of verbs, V15, which require a human or human-like noun in the position of N1 : (23) *On a provoqud la table ~ une chute brutale A restriction then disallows la conversion.., in the position N1 of (21). 2. The semantic data.</Paragraph> <Paragraph position="9"> From the analyses presented above, we see that two distinct types of semantic data are incorporated in the grammar.</Paragraph> <Paragraph position="10"> (1) A subclass of some major grammatical class cannot appear in a given position. This was the case for the verb subclasses V15 and V16, which require a human noun in one of the positions of their NP N object; for the noun object of ~tre, which can drop the article only if the noun is in the subclass N17; and for the object of verbs like confirmer, ddmontrer, ... which cannot be a concrete noun. (2) Some sequence of conjoined strings is not possible for given values of one of the strings or of some subclass appearing in the strings. Thus, the two conjoined nouns in the sequence (sentenceJM-1) depuis N1 et N~ may both be nominalizations, time nouns, or concrete nouns, but not one of each kind.</Paragraph> </Section> <Section position="6" start_page="26" end_page="26" type="metho"> <SectionTitle> 404 MORRIS SALKOFF </SectionTitle> <Paragraph position="0"> Another example was seen in the sentence JM-4 where the conjunction of dissimilar objects of a verb is ungrammatical. These two examples illustrate the problem of the conjunction of classes and strings which is not solved for the general case.</Paragraph> <Paragraph position="1"> The interesting point is that these two types of semantic restrictions are of the same nature as the syntactic restrictions and are incorporated in the grammar in the same way as the latter. This means that there is no need for a semantic component, or for semantic considerations that are completely separate from the usual grammatical procedures. The definition of subclasses is required in any case by the grammar, e.g.</Paragraph> <Paragraph position="2"> for the syntactic subclasses (singular, plural .... ); and the specification of the conjoinability of given sequences (strings) is required in order to treat conjunctions. In this way, the semantic component becomes part of the syntax and is incorporated without any special mechanism, The same type of restriction as that which forbids *Pierre sont ici or *L'homme est courageuse is used to prevent the analyzer from presenting a sequence such as *provoquer la table ~ la violence.</Paragraph> <Paragraph position="3"> Hence I extend this use to semantic subclasses. Some of these sub-classes can be defined syntactically, e.g., N17, but in any case they are sometimes used to exclude sequences that are not necessarily syntactically forbidden. These sequences do not violate any rules of the grammar, but violate what is usually called a selection rule or a semantic constraint.</Paragraph> <Paragraph position="4"> To the extent that these semantic constraints, or selection rules, can in fact be formulated, the formulation can probably always be stated in one of the ways (1) or (2) given above. If this is true, then the syntactic analyzer based on string grammar which I have presented here can incorporate semantics as well as syntax. Two difficulties in this formulation via subclasses immediately present themselves: a) The subclass itself is difficult to define, e.g., the subclass human or human-like which is needed to define V15 and V16, or the subclass ~ concrete ~ noun.</Paragraph> <Paragraph position="5"> b) The decision as to whether a given word does in fact belong to some subclass is not always easy to make. In the discussion of JM-2, I said that apjoorter belongs to the subclass V15, and this is how apporter is presendy coded in the dictionary. There are however well-formed sentences in which the N1 P N~ object of apporter has a non-human noun in the position N~: (24) J'apporterai une solution au probl~me</Paragraph> </Section> <Section position="7" start_page="26" end_page="26" type="metho"> <SectionTitle> ON USING SEMANTIC DATA IN AUTOMATIC SYNTACTIC ANALYSIS 405 </SectionTitle> <Paragraph position="0"> but this seems to be limited to pairs of nouns standing in some relationship to each other, as solution and probl~me. This relationship is very hard to define, since other pairs of nouns, seemingly related in a similar fashion, do not yield well-formed sentences with apporter: (25) ,J'ai app0rtb un pied ~ la chaise But these difficulties are not specific to the analyzer nor to the string grammar that I use; rather, they are independent of the parsing strategy - no matter what the analyzer - and will be solved, if indeed they can be solved, by more detailed research into the linguistic problems involved. From a practical point of view, this uniform treatment of the semantic and the syntactic data leads to a more compact granunar and a simpler analyzer than one containing separate semantic and syntactic components. Only semantically and syntactically correct analyses are furnished by such an analyzer, as is desired, and this is of prime importance for later applications of the analyzer to the problems of automatic translation or automatic documentation.</Paragraph> </Section> <Section position="8" start_page="26" end_page="26" type="metho"> <SectionTitle> 406 MORRIS SALKOFF MI~CANISMES BIOCHIMIQUES ET GI~N~TIQUES DE LA R.~GULATION DANS LA CELLULE BACTI~RIENNE </SectionTitle> <Paragraph position="0"> par Francois Jacob et Jacques MONOD.</Paragraph> <Paragraph position="1"> Services de Gen~tique microbienne et de Biochimie cellulaire, Institut Pasteur, Paris. I. INTRODUCTION Les connaissances acquises depuis quelques ann~es sur la structure des macromol~cules biologiques essentielles, acides nud~iques et prot~ines, permettent de comprendre, au moins dans ses grandes lignes, le rapport entre les fonctions de ces macromol~cules et leur structure chimique. L'&ude de la r~plication de I'ADN in vivo et in vitro a apport6 la preuve que le m~canisme chimique fondamental de l'h&~dit~ est bien celui qu'avaient propos6 WATSON ct CatCK (1953) ~. la suite de leur d6couverte de la structure de I'ADN. La d~couverte de I'AKN messager et de son r61e dans la biosynth~se des prot~ines, l'&ude des processus de transcription, les recherches sur le d&erminisme g6n&ique des structures primaires des prot6ines ont enti~rement confirm4 en les renouvelant, les anciennes hypotheses sur les relations entre les d~terminants g~n&iques et les prot6ines-enzymes.</Paragraph> <Paragraph position="2"> Les progr~s de la biochimie r6alisds depuis 50 ans ont, en outre, d~montr~ que les propri&~s, les structures, les activit~s qui caract&isent une cellule doivent, en dSfinitive, &re rapport~es ~ la structure et h l'activit~ des prot~ines que cette cellule est capable de synth&iser. Or, la structure de chacune de ces prot~ines est int6gralement d~finie, on peut 1'affirmer aujourd'hui, par un segment g6n&ique. Mais alors que la structure de I'ADN d'une cellule ou d'une lign6e cellulaire est invariante, les propri&~s biochimiques r&lis~es et exprim~es par cette cellule pourront ~tre profond~ment diff&entes; en outre, ces propri&~s sont modifiables en fonction de signaux chimiques sp~cifiques venus de l'ext&ieur. L'injection d'hormone thyroidienne ~ un t&ard provoque, avant m~me toute manifestation morphologique, la conversion de son syst~me excr&oire qui, de semblable ~ celui d'tm poisson, deviendra chimiquement analogue ~, celui d'un mammif~re: quelques heures apr~s rinjection, hs enzymes sp& cifiques du cycle de l'ur~e augmenteront en proportion de 50 ~ 100 lois et l'animal excr&era de l'ur~e alors que, jusque-l~t, il n'avait excr6t6 clue de l'ammo-BULL. SOC, CHIM. BIOL., 1964, 46, N deg 12.</Paragraph> <Paragraph position="3"> 1. PHRASE = 2. CHAINE D'ASSERTION = 3. GN = 4. DN = 5. A S DE V OMEGA = 6. VE OMEGA-PASSIF = 7. CHAINE D'ASSERTION = 8. PN = 9. PN =</Paragraph> </Section> <Section position="9" start_page="26" end_page="26" type="metho"> <SectionTitle> INTRO CHAINE CENTRALE MARQUE DE FIN 2. * (PARTJ SUJET *(PART.) VERBE * 3. CONNAISSANCES 4. PERMETTENT OBJET DV * 5. ARTICLE QUANT ADJ LES VE OMEGA-PASSIF 6. CHAINE D'ASSERTION </SectionTitle> <Paragraph position="0"/> <Paragraph position="2"> 0. RAPPORT 11.</Paragraph> <Paragraph position="3"> G.P. PREPOSITION D.P. N DEPUIS 12. ANNE1ES 13. G.P. PREPOSITION D.P. N 1. PHRASE = INTRO CHAINE CENTRALE MARQUE DE FIN 2.</Paragraph> <Paragraph position="4"> 2. CHAINE D'ASSERTION =* (PART.) SUJET * (PART.) VERBE * 3. CONNAISSANCES 4. PERMETTENT OBJET DV * 5.</Paragraph> <Paragraph position="5"> 3. GN = ARTICLE QUANT ADJ LES 4, DN = VE OMEGA-PASSIF 6.</Paragraph> <Paragraph position="6"> 5. A S DE V OMEGA --- CHAINE D'ASSERTION ET 7. ET 8, 6. VE OMEGA-PASSIF = VE * OM-PASS DV * ACQUISES 9, 7. CHAINE D'ASSERTION = * (PART.) SUJET * (PART.) VERBE * D- 23. MACROMOLE1CULES BIOLOGIQ VIRGULE UES ESSENTIELLES , 24.</Paragraph> <Paragraph position="7"> 22. GN = ARTICLE QUANT ADJ CES 23. GN = ARTICLE QUANT ADJ -ES 24. CONJONCTION = GROUP NOM ET ACIDES NUCLEI lQUES ET 25. 25. CONJONCTION = GROUP NOM PROTEIlNES, *** PLUS DE PLACE DANS LES ARCHIVES *** Fig. 3.</Paragraph> <Paragraph position="8"> ON USING SEMANTIC DATA IN AUTOMATIC SYNTACTIC ANALYSIS 411 TEMPS D ANALYSE = 1359 SEC/100 JM-1 *************************** NO 3 .................... LES CONNAISSANCES ACOUISES DEPUIS QUELQUES ANNE1ES SUR LA STRUCTURE D- -ES MACROMOLE1CULES BIOLOGIOUES ESSENTIELLES , ACIDES NUCLEIlQUES ET PROTEIlNES , PERMETTENT DE COMPRENDRE , A- -U MOINS DANS SES GRANDES LIGNES , LE RAPPORT ENTRE LES FONCTIONS DE CES MACROMOLE1CULES ET LEUR STRUCTURE CHIMIQUE . 1. PHRASE = 2. CHAINE D'ASSERTION = 3. GN = 4. DN = 5. CONJONCTION = 6. A S DE V OMEGA = 7. VE OMEGA-PASSIF = 8. CONJONCTION = 9. CHAINE D'ASSERTION = 10. PN = 11. PN = 1. PHRASE = 2. CHAINE D'ASSERTION = 3. GN = 4. DN = 5. CONJONCTION = 6. A S DE V OMEGA = 7. VE OMEGA-PASSIF = 8. CONJONCTION = 9. CHAINE D'ASSERTION = G.P. PREPOSITION D.P. N DEPUIS 16. ANNE1ES 17.</Paragraph> <Paragraph position="9"> G.P. PREPOSITION D.P. N A- -U MOINS DANS 18. LIGNES ARTICLE QUANT ADJ LE G.P. PREPOSITION D.P. N ENTRE 19. FONCTIONS 20.</Paragraph> <Paragraph position="10"> 16. GN 17. PN 18. GN 19. GN 20. PN 414 MORRIS SALKOI~F 15. CHAINE D'ASSERTION = * (PART.) SUJET * (PART.) VERBE * OBJET 1. PHRASE = INTRO CHAINE CENTRALE MARQUE DE FIN 2.</Paragraph> <Paragraph position="11"> 2, CHAINE D'ASSERTION = * \[PART.\] SUJET * \[PART.\] VERBE * OBJET 3. E1TUDE 4. A APPORT DV * E1 5. PREUVE 6.</Paragraph> <Paragraph position="12"> 3. GN = ARTICLE QUANT ADJ LA 4. NOM COMP = PREPOSITION GROUPE NOM DE 7. RE1PLICATION 8. IN-VlVO E 5. GN = 6. DN = 7. GN = 8. NOM COMP = 9. CONJONCTION = 10. QUE C1/C15 = 11. GN = 12. CHAINE D'ASSERTION = 13. GN 14. NOM COMP</Paragraph> </Section> <Section position="10" start_page="26" end_page="26" type="metho"> <SectionTitle> RELATIONS ENTRE LES DE1TERMINANTS GE1NE1TIQUES ET LES PROTEIlNES - ENZYMES </SectionTitle> <Paragraph position="0"> 1. PHRASE = INTRO CHAINE CENTRALE MARQUE DE FIN 2.</Paragraph> <Paragraph position="1"> 2. CHAINE D'ASSERTION = * (PART.) SUJET VIRGULE * (PART.) 3. GN 4. NOM COMP 5. PN 6. CONJONCTION 7. VE OMEGA 8. GN 9. CONJONCTION 10. GN 11. NOM COMP 12. GN 13. NOM COMP 14. CONJONCTION 15. CS3 VANT OMEGA 3. DE1COUVERTE 4. 5. , 6. ONT 1. PHRASE = INTRO CHAINE CENTRALE MARQUE DE FIN 2.</Paragraph> <Paragraph position="2"> 2. CHAINE D'ASSERTION = (PART.) SUJET (PART.) VERBE 3. PROGRE2S 4. 5. ONT , EN O OBJET D.V.</Paragraph> <Paragraph position="3"> UTRE, 6.</Paragraph> <Paragraph position="4"> 3. GN = ARTICLE QUANT ADJ LES 4. NOM COMP = PREPOSITION GROUPE NOM DE 7. BIOCHIMIE 5. DN = VE OMEGA-PASSIF 8.</Paragraph> <Paragraph position="5"> 6. VE OMEGA = VE OBJET D.V. DE1MONTRE1 9.</Paragraph> <Paragraph position="6"> 7. GN = ARTICLE QUANT ADJ LA 8. VE OMEGA.PASSIF = VE OM-PASS D.V. RE1ALISE1S 10.</Paragraph> <Paragraph position="7"> 9. QUE C1/C15 = QUE CHAINE CENTRALE OUE 11.</Paragraph> <Paragraph position="8"> 10. P N = G.P. PREPOSITION D.P. N DEPUIS 12. ANS 11. CHAINE D'ASSERTION -- (PART.) SUJET VIRGULE (PART.) VER 13. PROPRIE1TE1S , 14. DOIVENT , BE OBJET D.V. *</Paragraph> </Section> class="xml-element"></Paper>