<?xml version="1.0" standalone="yes"?>
<Paper uid="C86-1032">
  <Title>COMPUTATIONAL COMPARATIVE STUDIES ON ROMANCE LAGUAGES A linguistic comparison of lexicon-grammars</Title>
  <Section position="1" start_page="0" end_page="0" type="metho">
    <SectionTitle>
COMPUTATIONAL COMPARATIVE STUDIES ON ROMANCE LAGUAGES
</SectionTitle>
    <Paragraph position="0"> A linguistic comparison of lexicon-grammars Annibale ELIA &amp; Yvette Mathieu Istituto di Linguistica Universita di Salerno Laboratoire d'Automatique Documentaire et Linguistique C.N.R.S. - Universit~ de Paris 7 I. The linguistic comparison of lexicon-grammars What we present here is an application on the basis of the italian and french linguistic data bank assembled by the Istituto di Linguistica of Salerno University (Italy) and the Laboratoire Automatique Documentaire et Linguistique (C.N.R.S,-France). These two researching centers have been working for years to the constitution of formalized grammars of the respective languages. The composition of lexicon-grammars is the first stage of this project.</Paragraph>
    <Paragraph position="1"> A lexicon-grammar requires the study of nearly 300 syntactic property for all words of a language.</Paragraph>
    <Paragraph position="2"> The data revealed in this way are memorized under a matricial form of the type</Paragraph>
    <Paragraph position="4"> Therefore we have a number of classes of nouns, adjectives, adverbs, conjunctions and verbs, the determiners and the prepositions being included in the verbal classes.</Paragraph>
    <Paragraph position="5"> Some matrices exist with lexicon-grammars correspondances between two lexicon-grammars as:</Paragraph>
  </Section>
  <Section position="2" start_page="0" end_page="148" type="metho">
    <SectionTitle>
TABLE 45
</SectionTitle>
    <Paragraph position="0"> - - - a~14ognarm - - - + - + + - - + + + + .... / .... +volt beloin - ibu*~. - - + - + .... / / + * .... + / - / - *bul,~ e' - *pp~o*ltta,* - - * / + / - + - + + + * .... + .... pro+it.~ e -- - apprQ@ltta~sl - - * + + * = - - + * * + * - _ + + + - _ - en prDfJter - ~lfiid*r. + - + / - + * + - + + * + .... + - - / - ~* *l~l.r - di**ntlC/*~l - - + - + - + * + + * + * - - - + + - - + / aubillr 6 . - - dlspararl - + / * + + + + - + + - / .... + - . - + d4lilpOrlr e 4llp*~ar*i - - + + + + - / - / / + + - - + + + + + + + um di~ispirer ~ua*tar* - - / / + + + + + + * * * .... / .... do~t,~ B - - - isultarl + - + * + + - / - / * * * + - + * + * + + + exulter 8 flni*stlce~* - - + * + + + + + + + + + - + - - + + - / + rlve, B - - 'lavoiar. - - + + + - + + - * + * * - + + - + + - + + ~*ver - C/IvoligllIPl - - + + + + + + - + + + + .... + + - + * livlr B -'deg&lt;'.-... ....</Paragraph>
    <Paragraph position="1"> .... +,o,,, :::::::::::::::::::::: ..... .</Paragraph>
    <Paragraph position="2"> ? llpillirl - + + + - + - + - + + / + - - _ / ..... rlCfollr B + + - _ niClllttir I - - - + - + *+ * - . + + + .... + .... niCll.l~ir I0 - plnttril - - + + + + + + - + + + + - - + + + - _ + + ii rlplntlr 17 + + - - P coitttlPl - - + + + + + + - * + + + + - - + + .... proil+lr B -- +lcor+irl t - - + - + - + + + + + + + - + - + + - . + * ii pip+ill+ m - rldlPllnl - - + * + + + i - * * + + - - - + ..... ll rlr+ B + - _ _ Pill~iirlt : + + + + + - + - + + + + - - _ . + + + + + ii tilllntl+ m + - = - .C/.l+arsl - - + - * - * * + * * * + - - - + * - % + + ++bilep 6 - .oi.*~. + - - * - * + + - + + + + - - - + + * - - - +i+.P , - IOV iv .iPll - - + + i - # @ + + + i + .... + - _ + + ii ~ipplilP 0 + - - - tac.~l - - * + + + - + + + + + + - + + + + + - * + +1 tilt, n - vlll+~n4p$i - - # i # i + + - # + + i i - + # # # # + + iioIp ~o~+i  Fig. 2 The composition of a lexicon-grammar allows to take into account selection and sub-categorization restrictions as well as derivational morphologic relations, without getting away from the limits of the formal syntax, that is, in limiting to the only experience of the estimation of acceptability, the recourse to intuitions of semantic nature.</Paragraph>
    <Paragraph position="3"> A lexicon-grammar can satisfactorily render an account for the lexical and syntaxical differences of a word and those morphologically connected to it, allowing as well a comparison with verbs and nouns of another language disposing of a lexicon-grammar.</Paragraph>
    <Paragraph position="4"> The high level of lexicon syntactic informations of the italian and french lexicon-grammars allowed us to a very subtle comparative study of the two languages. For exemple, the comparison of the italian verb investire and the french verb investir, leads to the following situation:</Paragraph>
    <Paragraph position="6"> (1) b. II governo ha investito una commissione parlamentare del compito di indagare sul caso Moro.</Paragraph>
    <Paragraph position="7"> INVESTIRE 2 (2) NO V Nl in N2 =: (2) a. Max ha investito un milione nell'affare (2) b. Eva ha investito (molto, tutta se stessa, il suo affetto) nel rapporto con Max INVESTIRE 3 (3) NO V N1 (E,con N2) -: (3) a. II generale investi la piazzaforte (E, con le truppe d'assaIto) (3) b. Luca invest~ (Eva, l'auto di Eva) (E, con l'automobile) (3) c. Eva invest~ Max (E, (con,di)(tutti i suoi problemi, parolacce)) INVESTIRE 4 (4) NO si V di N1 =: (4) a. Lotario si invest~ de\] potere regale INVESTIRE 5 (5) NO si V di Nl =: (5) a. Max si @ investito de\] (ruolo, personaggio) di Don Giovanni INVESTIRE 6 (6) NO V N1 =: 16) a. Questa diffic~}e scelta investe tutti noi (6) b. La trattazione investe anche i pronomi INVESTIRE 7 (7) NO V Prep NI =: (7) a. La nave ~ investita (contro~ su) Io scoglio  (3) NO V NI (E, avec N2) =: (3) a. Le g@n~ral a investi la ville {E, avec ses troupes)  The french correspondents for the other uses of investire are:  INVESTIRE 3b. renverser, heurter, entrer en collision INVESTIRE 3c. couvrir INVESTIRE 4 s'approprier, usurper INVESTIRE 5 entrer dans le r6le de INVESTIRE 6a. toucher, concerner INVESTIRE 6b. concerner INVESTIRE 7 6chouer  To give a comparative presentation of the italian and french data, we have two comparative tables of the realization of zhe morpheme INVEST- in izalian and french. The first one (fig. 3) gives a survey of the correspondance, while a detailed comparison of the distributional syntactic properties and of the paraphrases is given by the second one (Fig. 4) \[ N V E S T - null -IRE (1) ,,.</Paragraph>
    <Paragraph position="8"> -. 1 TIIRA f - ITOI~\] (I) c, -IRE (1) b.</Paragraph>
    <Paragraph position="9"> (1) d.</Paragraph>
    <Paragraph position="10"> -IRE (2) a. I - 12{ENTO i  This table is to be read from the top left where the verbs investire and investir are placed. In the middle the properties are put in a line and noted by indications placed at the right concerning the number and the kind of use (example, (a)... I. (b) .... 3, etc) At the top right, there are specifications on the support verb (Vsup) and the nominalization (V-n). The equivalent structures, italian-french, for prepositions and support verbs have a representation with a slash / (example di/de: di-italian and de-french; dare/donner: dare-italian and donner-french). At the left, there are two columns with plus and minus (+, -) which point out whether the property in question is valid or not for the verb; if the point is about the property with Vsup, first of all you must control if the verb at the left has a + or a - in the division of the column, if it is +, you can then read at the top right the type of Vsup and of V-n; if it is - , but there is a + at the right, that means that it refers only to one of the two verbs. The $ symbol is equivalent to &amp;quot;financial investments&amp;quot; and the symbol to &amp;quot;psychological investments&amp;quot;. null  Comparison between lexicon-grammars of different languages leads to draw up formal correspondences between structures, where so-called &amp;quot;grammatical words&amp;quot; are constants. Therefore, those &amp;quot;grammatical words&amp;quot; are considered as formal parts of structure. With that clause case, for example, we will have to take into account the following structures: italian french ...il fatto che F \]e fait que P ce que P ...che F que P which, will lead us to say that there is no italian correspondent to the french sentence  (1) Max doute de ce que Marie vienne in view of the fact that the italian sentence (2) *Max dubita di ci6 che Maria venga is unacceptable. But the sentences (3) Max doute du fait que Marie vienne {fr) (4) Max dubita del fatto che Maria venga (it) (5) Max doute que Marie vienne (fr) (6) Max dubita che Maria venga (it) drive us to consider italian structures (4) and (6) at a formal proximity degree of french structure (1).  From a strictly formal point of view, the correspondence between (1) and (2) cant'be accepted, but from a larger point of view, which implies the formal proximity degree notion (obviously depending on the researcher's theoretical choice) we can mention structures (4) and (6) as the closest realization of the french structure (1). Of course, strucLures (4) and (6) have the strict formal correpondents (3) and (5). The very complex data organization is usable by specialists but not by ordinary users. That is why we have been interested by the realization of a flexible and intelligent system, which allows the consultation and comparaison of the two languages lexicon-grammars (which are at the present day romance languages lexicon-grammars), and which includes the correspondance notion we have just presented. Its use will not need any knowledge neither of the data organization nor of computer science.</Paragraph>
  </Section>
  <Section position="3" start_page="148" end_page="148" type="metho">
    <SectionTitle>
2. The TRANSLEGS system
</SectionTitle>
    <Paragraph position="0"> Fig. 5 A &amp;quot;query-answer&amp;quot; system of the interface allows Lo choose the type of research one wants to make. The system offers a research arborescence guided by its own resources and by the user's choices.</Paragraph>
    <Paragraph position="1"> If the study concerns, for example, a class drawn up a know definitional property, TRANSLEGS will look for all the properties of this class, whether the correspondents exist in the other language, which elements appear (nouns, verbs, adjectives), in which struc~ures do they appear and how do the correspondents of the other language behave.</Paragraph>
    <Paragraph position="2"> If the study concerns a particular verb or noun, the system will look for the strucLures where this element appears, it will find its correspondent in the other language and the classes in which it exists, its properties and those of the correspondent into a given class. It will look for elemenLs that have a degree of proximity for the secondary corresponden~ ce, and how do they behave in front of the properties and associated sentences.</Paragraph>
    <Paragraph position="3"> in some cases, TRANSLEGS's answer needs deduction rules, if properties or classes exist in one language and not in the other one for example, fhen, the system will propose &amp;quot;substitution&amp;quot; classes according to their formal proximity degree, lh# same type of answer will be done if one element has some properties in one language that its correspondent doesn't have.</Paragraph>
    <Paragraph position="4"> The system being quite supple, the user can create his own environment a) by modifying the deduction rules b) by building new structures according to linguistic choices different from those reflected by existing classes (in that cas TRANSLEGS will make a class from the choosen definition property and those linked to it).</Paragraph>
    <Paragraph position="5"> The system is written in PROLOG and runs on a VAX ll/  750 from Digital Equipment.</Paragraph>
    <Paragraph position="6"> 3. Examples of dialogues  In the following examples, TL means TRANSLEGS  correspondante.</Paragraph>
    <Paragraph position="7"> Ex. 3 ammirazione avec avere a la propri6t6 D#t = 0 admiration ne l'a pas Suite propos6e des propri#t~s correspondantes les plus proches:</Paragraph>
    <Paragraph position="9"/>
  </Section>
class="xml-element"></Paper>