<?xml version="1.0" standalone="yes"?>
<Paper uid="C88-2154">
  <Title>DLT - AN INDUSTRIAL R &amp; D PROJECT FOR MULTILINGUAL MT</Title>
  <Section position="6" start_page="756" end_page="759" type="concl">
    <SectionTitle>
4. Project outlook
</SectionTitle>
    <Paragraph position="0"> BSO is now in the process of preparing for the next phase of the project (budgeted at 12 million guilders), in which the emphasis will be on large-scale dictionary and knowledge-base expansion, and relaxation of grammar restrictions.</Paragraph>
    <Paragraph position="1"> Work done on DLT in the past 5 years confilans the feasibility of its architecture and its instrumental use of Esperanto. Some of the modifications to Esperanto thought necessary \[Witkam, 1983\] in its pivotal MT function appeared to be unnecessary as work on DLT progressed, i.e. the Intermediate Language is closer to Esperanto now than it looked like in the beginning. Criticizers' predictions that the ill_, would keep changing and would drift further and further away from Esperanto, have not been borne out.</Paragraph>
    <Paragraph position="2"> The essence of DLT is not so much an attempt at an unambiguous intedingua, but rather: a split of the overall ITanslation sequence into a form part and a ~ part, in which file former is arranged as a .~ouLb.~.:dir_LC/_~ and the latter as an interlingual /VII&amp;quot; pl'oc;ess., The doublerdirect process is the surface translation referred to above (Steps 2 and 5), which could be loosely characterized as &amp;quot;.dumb syntax&amp;quot;. The interlingual process corresponds Io Steps 3 and 6 above, and contains all the 8e_r~tics (including knowledge-based inferencing).</Paragraph>
    <Paragraph position="3"> Meanwhile, the share of IL-based or IL-directed work in the overall translation sequence - form part as well as content part - has increased considerably, compared to the initial design of 1982 \[see Fig. 2\]. In the form part, which is bilingual and purely syntactic, the Esperanto 1L plays the role of &amp;quot;metataxis partner&amp;quot; for every source and target language ('metataxis' is tile contrastive-syntactic transformation of dependency trees). The content part is monolingual, i.e. the semantics is a question of IL-only. It goes withont saying that such a design can only succeed by virtue of the fact that Esperanto is a well defined language on its own, with a well defined syntax and lexicon, and with the help of project staff fluent in or at least conversant with that language.</Paragraph>
    <Paragraph position="4"> The logistics of DLT development draw heavily upon the existence of Esperanto resources in the widest sense: linguists with Esperanto training, Esperantists with degrees in languages or with extensive translator's experience, corpora of moderu Esperanto texts, etc. Timely availability of these resources in sufficient quantity and quality demands some special organizational and promotional activity from the DLT entrepreneurs.</Paragraph>
    <Paragraph position="5"> This means crossing international borders, including the border between EEC and COMECON countries. East European countries have a relatively large base of Esperanto speakers, and much on-going activity. The Budapest Eotvos Lorand University has a chair in Esperanto. In Poland, a new Esperanto center connected with the University of Bjalistok has been founded in 1987. In Bulgaria an international training center exists, and in l~'ague an International Center for Esperanto Terminology has recently been established.</Paragraph>
    <Paragraph position="6"> The challenge of the DLT project is therefore as much an organizational as a technological challenge. Cooperation with Hungarian, Finnish and Czech linguists has already begun, and preparatory work has been arranged with collaborators in East Asian countries. Diversity of language types is a deliberate aim for the multilingual DLT system of the 1990s, and tile designed architecture makes it technically feasible. It can be hoped that, against the background of 'glasnost' and 'perestrojka', fruitful and effective East-West cooperation may add to the success of a system for international use and of general interest to the growing community of computational linguistics students and researchers.</Paragraph>
    <Paragraph position="7">  1982 and the 1987 designs of the DLT translation from source to target language (top-to-bottom). The shaded parts indicate where the Intermediate Language is involved.</Paragraph>
  </Section>
class="xml-element"></Paper>