<?xml version="1.0" standalone="yes"?>
<Paper uid="C80-1063">
  <Title>A MACHINE TRANSLATION SYSTEM FROM JAPANESE INTO ENGLISH -- ANOTHER PERSPECTIVE OF MT SYSTEMS --</Title>
  <Section position="8" start_page="422" end_page="422" type="concl">
    <SectionTitle>
5. Concluding Remarks
</SectionTitle>
    <Paragraph position="0"> Fig. 17. shows some examples of translation which illustrate the current abilities of the system.</Paragraph>
    <Paragraph position="1"> As these examples show, the system can translate fairly complex sentences, though several problems still remain unsolved. The distinction between definite and indefinite noun phrases, for example, cannot be made by the current system, because no fixed expressions to distinguish them exist in Japanese surface sentences. Therefore, neither definite nor indefinite articles are not attached to the English noun phrases. Another problem is to supply appropriate elements from context for omitted expressions. Especially, case elements in a sentence are frequently omitted in Japanese, when they are easy to recognize from the context. Though the current system tries to find appropriate surface English words and structures at the English synthesis phase which do not require the omitted elements, it would be inevitable to incorporate contextual processings. The current system works very well as an experimental prototype. Following to the same basic principles with the current system, we are now designing a new and more advanced ~--~ system, in which these (i) Input Japanese ~ 0 ~ ~ ,~ defects of the current system will be improved.</Paragraph>
    <Paragraph position="2"> Our basic contention in this paper is that most of linguistic phenomena should be treated by understanding systems, including the authors,emphasized too much the importance of pragmatic knowledge. However, one of the recent trends in this area, which we also support, is to lay more emphasis on the importance of syntactic processings, or at least, syntactic structures of sentences. This attitude is, we believe, especially important for MT systems. The various transformed syntactic structures described in section 2 have been overlooked by the researchers of computational linguistics so far. We hope that our approach, the lexicon based analysis procedure, provides an appropriate framework to integrate syntactic structures and operations with the other kinds of processings such as semantic and pragmatic ones.</Paragraph>
    <Paragraph position="3"> Result of Translation User header label and user *railer labe |can be written out to sequential da %a ~ e*.</Paragraph>
    <Paragraph position="4"> lexical rules, instead of general syntactic rules..z *** This leads us to the framework called lexicon ~, I~ po6slble ~o write Out user header labo\] and user trailer label to based procedures. This seque.*lal d~*~ ~*.</Paragraph>
    <Paragraph position="5"> approach is not only fairly compatible with D D ~ ~ D S N A M E p~ 9 ~ P It. ~ w /~ ~ ~ ~ ~ b te ~ ~ . the recent trends inlin- (2) Input Japanese ~ ~ 'J ~ ~ ~I~ la N A M E ~ ~ / ~' ~'~ t~ ~ ~ ~f N A M guistics, but also gives E ~-&amp;quot; 9 J ~ la ~ ~ -r ~ ZZ ~ PSt ~ J, w /~ ~ ~' ~ ~ ~ ~ o us a good framework in which grammars can be Result of Translation easily revised and augmented by modifying the ~* *** lexical description of each individual word, Whe. member name was spec, fled In DSNAME paramelep of DD .~.ta+emeR*, ,'.f without any modifications NA&amp;quot;E parame%er Is spec Irked In utl li %y control stalem~nt , member name of the general framework.speC/~,~ b~ NAME paramete~ takes peeceda, ce . The next comment is about =z *** the relationships among syntactic, semantic, and When member name was specl fled I n DSNAME parameter of DD ~a *ement , If pragmatic processings. NAME paramo%er Is specified |n utl II *y control statemer~t , member name ~hlch At the early stage of is Bpecl fled by NAME parame%er takes pPecedance. development, the research- Fig. 17. Translation Results ers of natural language ---423--</Paragraph>
  </Section>
class="xml-element"></Paper>