File Information

File: 05-lr/acl_arc_1_sum/cleansed_text/xml_by_section/metho/80/c80-1025_metho.xml

Size: 6,700 bytes

Last Modified: 2025-10-06 14:11:20

<?xml version="1.0" standalone="yes"?>
<Paper uid="C80-1025">
  <Title>PIIOBLi~IS OF F01@biL REPRESENTATION OF TEXT STRUCTURE FROM THE POINT OF VIEL{ OF AUTOMATIC TitiNSLATION</Title>
  <Section position="1" start_page="0" end_page="174" type="metho">
    <SectionTitle>
PIIOBLi~IS OF F01@biL REPRESENTATION OF TEXT STRUCTURE
FROM THE POINT OF VIEL{ OF AUTOMATIC TitiNSLATION
Z.M. Shalyapina
</SectionTitle>
    <Paragraph position="0"> Institute of 0rien~al Studies of tile Academy of Sciences of tile USSR l.loscow, USSR Stumnary The paper is devoted to linguistic problems of defining the basic formalized representation of text in an automatic translation system within the frsmlework of the so-called integral formal model of tile tr~ulslation process, the primary requirement for tills representation considered to be a compromise between its semal~ticity, superficiality~ a/id exhaustiveness. A representation covering five major aspects of text structure (its lexico-gra~\]matical composi$ion; its predicate-ar~iment organization on the semantico-syntactic level; the syTitactic groupinz ~ of its units; the a/iaphoric relations between them; tile peculiarities of their linear arrmlgement) and referred ~o as Combined Structural Representation (CSR) of text, is described to show tile ways and means of achieving this compromise in the Japanese-Russian Automatic Translation Project, now under development at tile Institute of Oriental Studies of the Academy of Sciences oPS the USSR (Moscow).</Paragraph>
    <Paragraph position="1"> Introduction Many problems of the automatic processing of text require for their effective solution a previous analysis of the text processed, aimed at tra/isforming this text into its intelnnediate formalized representation of some kind, more suitable for further processing than the text itself. I'~hen determining tile concrete characteristics of such a representation one must obviously take into account the operations meant to be applied to it, or to be performed on its basis within the framework of the system ilivolred. If it is the problem of automatic tra~Islation that the system is to solve, the set of tile corresponding operations will depend primarily on the general formal model of the translation process underlying this system. One version of tlle model in question, proposed in I a/id discussed in more detail in 2, envisages the following main groups of operations: 1 ) analysis and interpretation of the initial text, simulating the process of perceiving and understanding its signification and denotation; ideally, it presupposes a semantic description of tile text, as well as a model of the situation (&amp;quot;world&amp;quot; fra~ment) presented in it, being&amp;quot; constructed from this text (possibly, via a ,lumber of intelnnediate representations ) ; 2) tra/islation proper, which is performed at a level R of some formal representation R i of the initial text, derived from its analysis, and 6ullounts to selecting translation equivalents for the units included in l{i: the result is an intermediate representation R t of tlle target text, this representation being usually (although not necessarily) of the same level as Ri; 3) verification of the adequacy of the tr~ulslation performed, by means of a/lalyzili~' the resulta/it representation R t a/id comparing the semantic description and the situational model obtained, with the sema/itic description alid the model of tlre situation corresponding to the initial text.; I~) generation (synthesis) of the tar,z'et text by tra/lsforming tlre inte~nediate representation R t formed during translation proper and assumed to be adequate by the verification procedure, into a sequence of actual word-forms azld punctuation marks making up the tartset l.a/i~age text; 5) evaluation of ~he target text with a view fie detect undesirable ambi-Ln/ities and inaccuracies that mi~-ht have slipped in during the synthesis process; it implies 8/lalyzin~- the text back to the I{ level and checking,whether tile resulting representation R t coincides with the representation R t from which this text has been formed; 6) editing operations dictated by the checks and comparisons made:if the translation is judged to be inadequate tliey will consist ill returning&amp;quot; to the phase of translation proper and either substituting alternative translation equiva- , lents for some of the previously selected ones,or reconsidering the entire procedure %Ised and repeating it at a different (&amp;quot;deeper&amp;quot;) representation level or in a different form (probably,resorting * to synonymous trsl\]sformations of&amp;quot; the initial text at the R i level); if it is tile target text ambif~lities 6uld stylistic imperfections that arc ~o be removed,better expressive means will be sought chiefly by actuating9 the system of synonymous tra/Isformations at the R t level.</Paragraph>
    <Paragraph position="2">  It is readily seen that the basic level of formal text representation from the standpoint of the above conception of the translation process is level R, directly concerned with the most important translation operations, primarily, the operations of translation proper, the scope of ~lich is practically confined to the level in question, and the operations of synthesis ensuring the transition from the R-level representation of a text to its more &amp;quot;superficial&amp;quot; representations up to the text as such.</Paragraph>
    <Paragraph position="3"> Some other of the operations mentioned involve also switching from the R-level to &amp;quot;deeper&amp;quot; levels of intermediate formal text representation and taking into consideration such supplementary factors as the essence of the situation described by the text to be translated, the semantic peculiarities of the vocabulary and the syntax of the two languages; the requirements of gr~nmaticality and stylistic normativity (regularity) of the target text, and so on. The foregoing shows that these operations are mostly auxiliary in nature, their main purpose being to improve the content adequacy and the linguistic acceptability of the translation text formed through the use of the R-level representation; in a concrete automatic translation system based essentially on ~he formal model we have outlined, they may be reduced or even altogether omitted for various practical reasons.</Paragraph>
    <Paragraph position="4"> However, whether these supplementary operations be included in an AT system or not, it is clear that the system will depend largely for its efficiency on the choice of the intermediate level R. It is precisely this basic level that we are now Going to consider.</Paragraph>
  </Section>
class="xml-element"></Paper>
Download Original XML