<?xml version="1.0" standalone="yes"?>
<Paper uid="W01-1507">
  <Title>Multilingual ISLE Lexical Entry</Title>
  <Section position="5" start_page="0" end_page="0" type="evalu">
    <SectionTitle>
3 Methodological and organisational
</SectionTitle>
    <Paragraph position="0"> issues As in previous EAGLES, it is considered helpful to base the recommendations on the requirements stemming from a few application systems. The CLWG agreed to focus on two major broad categories of application: machine translation (MT) and cross-lingual information retrieval (CLIR).</Paragraph>
    <Paragraph position="1"> As said above, the CLWG has agreed that we should base any multilingual description on monolingual descriptions. MILE should therefore include previous EAGLES recommendations for other layers. We must evaluate the usefulness of these layers with respect to multilingual tasks, focusing in particular on MT and CLIR tasks. Obviously an additional module is needed, where correspondences between languages are defined, including conditions on syntactic structures involving lexical entries. The linking module (transfer) may not be the same for different applications: it may be simpler for CLIR, which may be a subset of the one needed for MT. For CLIR, an ontology or semantic hierarchy is however required.</Paragraph>
    <Paragraph position="2"> We are also adopting an approach that would lead to a formalisation of the information contained in traditional bilingual dictionaries, such as restrictions on translation, collocations and examples.</Paragraph>
    <Paragraph position="3"> The CLWG agreed the following were appropriate tasks to concentrate on, in order to discover basic notions for MILE:  1. Analyse information given to the human user in bilingual/monolingual dictionaries that allows selection of correct equivalence. 2. Analyse (if these can be obtained) instructions/guidelines supplied to lexicographers for writing bilingual entries. 3. Investigate, in corpus concordances, which are the clues that allow to disambiguate/decide on proper sense for translation.</Paragraph>
    <Paragraph position="4"> 4. Elaborate a typology of transfer conditions/actions and investigate lexical requirements.</Paragraph>
    <Paragraph position="5"> 5. Look at multilingual lexical requirements  for approaches based on interlingual concepts/ontologies.</Paragraph>
    <Paragraph position="6"> 6. Rank our typology in terms of scale of difficulty of disambiguation 3.1. Types of information to be addressed Regarding the various types of information to be addressed, the following &amp;quot;workflow&amp;quot; was agreed:  and is not otherwise used in applications (e.g. collocation type), or there are notions from other layers that we have not already considered: * decide which method is needed to do work on it * prioritise: what is used already in multilingual lexicons (but not covered in EAGLES, e.g. covered in OLIF) and also then look at what needed in near future * record what needs further development. A starting point will be the previous EAGLES recommendations, as instantiated in PAROLE/SIMPLE, for which - as said above there is a unique DTD for all the 12 languages involved. This will be revised and augmented after work done on various types of information. ISLE will also implement a lexicographic tool, with which a sample of lexical entries will be encoded according to the MILE structure. Assignments for in-depth analysis of the information types were done, and work is now carried out by the various CLWG members. Results of on-going work will provide: (i.) a list of types of information that should be encoded in each module; (ii.) linguistic specifications and criteria; (iii.) a format for their representation in multilingual lexicons; (iv.) their respective weight/importance in a multilingual lexicon (towards a layered approach to recommendations).</Paragraph>
  </Section>
class="xml-element"></Paper>