<?xml version="1.0" standalone="yes"?>
<Paper uid="P98-1053">
  <Title>Accumulation of Lexical Sets: Acquisition of Dictionary Resources and Production of New Lexical Sets</Title>
  <Section position="4" start_page="332" end_page="333" type="metho">
    <SectionTitle>
3 Virtual Accumulation
</SectionTitle>
    <Paragraph position="0"/>
    <Section position="1" start_page="332" end_page="332" type="sub_section">
      <SectionTitle>
Abstraction of Lexical Sets
and
3.1 Virtual accumulation
</SectionTitle>
      <Paragraph position="0"> Accumulation discussed so far is real accumulation: the LS acquired or produced is available in its whole and its elements are put in a &amp;quot;standard&amp;quot; form used by the lexical system. However, accumulation may also be virtual, ie LSs which are not entirely available may still be used and even integrated in a lexical system, and lexical units may rest in their original format and will be converted to the standard form only when necessary. This means, eg, one can include in his lexical system another's Web online dictionary which only supplies an entry to each request.</Paragraph>
      <Paragraph position="1"> Particularly, in virtual acquisition, the resource is untouched, but equipped with an acquisition operation, which will provide the necessary lexical units in the standard form when it is called. In virtual production, not the whole new LS is to be produced, but only the required unit(s). One can, eg, supply dynamically German equivalents of an English word by calling a function looking up English-French and French-German entries (in corresponding dictionaries) and then chaining them. Virtual production may not be suitable, however, for some production categories such as inversion.</Paragraph>
    </Section>
    <Section position="2" start_page="332" end_page="333" type="sub_section">
      <SectionTitle>
3.2 Abstraction of LSs
</SectionTitle>
      <Paragraph position="0"> The framework of accumulation, real and virtual, presented so far allows us to design a very general and dynamic model for lexical systems.</Paragraph>
      <Paragraph position="1"> The model is based on some abstraction levels of LSs as follows.</Paragraph>
      <Paragraph position="2">  (1) A physical support is a disk file, database, Web page, etc. This is the most elementary level. 2 UNL: Universal Networking Language (UNL 1996).  (2) A LS support makes up the contents of a LS. It comprises a set of physical supports (as a long dictionary may be divided into several files), and a number of access ways which determine how to access the data in the physical supports (as a database may have several index). The data in its physical supports may not be in the standard form; in this case it will be equipped with a standardizing function on accessed data. (3) A lexical set (LS) comprises a set of LS supports. Although having the same contents, they may be different in physical form and data format; hence this opens the possibility to query a LS from different supports.</Paragraph>
      <Paragraph position="3"> Virtual LSs are &amp;quot;sets&amp;quot; that do not have &amp;quot;real&amp;quot; supports, their entries are produced from some available sets when required, and there are no insert, delete activities for them.</Paragraph>
      <Paragraph position="4"> (4) A lexical group comprises a number of LSs (real or virtual) that a user uses in a work, and a set of operations which he may need to do on them. A lexical group is thus a workstation in a lexical system, and this notion helps to view and develop the system modularly, combinatorially, and dynamically.</Paragraph>
      <Paragraph position="5"> Based on these abstractions, a design on the organization for lexical systems can be proposed. Fundamentally, a lexical system has real LSs as basic elements. Its performance is augmented with the use of virtual LSs and lexical groups. A catalogue is used to register and manage the LSs and groups. A model of such an orgamzation is shown in the figure below.</Paragraph>
      <Paragraph position="6">  Although we have not yet been able to evaluate all the lexical data accumulated, our methods and tools for acquisition and production have shown themselves useful and efficient. We have also developed a rather complete notion of lexical data accumulation, which can be summarized as:</Paragraph>
    </Section>
  </Section>
  <Section position="5" start_page="333" end_page="333" type="metho">
    <SectionTitle>
ACCUMULATION = (REAL + VIRTUAL)
(ACQUISITION + PRODUCTION)
</SectionTitle>
    <Paragraph position="0"> For the future, we would like to work on methods and environments for testing accumulated lexical data, for combining them with data derived from corpus-based methods, etc. Some more time and work will be needed to verify the usefulness and practicality of our lexical system design, of which the essential idea is the combinatorial and dynamic elaboration of lexical groups and virtual LSs. An experiment for this may be, eg, to build a dictionary server using Intemet online dictionaries as resources.</Paragraph>
  </Section>
class="xml-element"></Paper>