Next: Basis for comparison
Up: Methodology Task
Previous: Introduction
- The EAGLES descriptive standards proposal is made for the
construction of a lexical resource, thought of as being
multifunctional, but not for a given application resource or for a
given range of applications. This has to be kept in mind in
discussions about generalization devices, about the use of
hierarchical or other types of structures superposed onto the EAGLES
descriptions, questions of redundancy or elimination of redundancy,
and so on. This is also related with questions of the formal
representation of proposals which will be the output of EAGLES work.
- The experience acquired in the different projects dealing with lexical
knowledge modelling has shown that it is not realistic to expect the
same degree of consensus for all levels of linguistic description. Typically,
while
arguments about orthography and morphology rest on well established
linguistic facts, arguments on semantics rest on theories for which no
experimentation has been extensive enough to make sure that the
descriptive devices used are already defined on the basis of sufficient
uncontroversial facts. For this reason, the methodology applied to
each level should be adjusted to the kind of results that it is likely
to produce.
The levels of linguistic description for which it is most
realistic to come up with proposals for descriptive guidelines in
the short or medium term should be considered with more emphasis
than the other ones; morphology and morphosyntax are among these
levels, see the revised workplan for that matter.
- It is not relevant to determine priorities on languages to define
standards for. The project will concentrate on overall features for
the languages of the EEC, while the possibility of representing other
languages - Swedish, for example - should not be excluded.
A comprehensive description of lexical features for all european languages is a
long task which would entail the participation of a specialist of each language
in each task dealing with each linguistic level. The EAGLES objective is
neither to provide such a comprehensive description nor to concentrate on
specific languages. Such language-oriented studies may lead to unsatisfactory
results such as, for example, a well developed morphology in German and French
and very little in English and Spanish while syntax would be completed only in
French and Italian according to the expertize available in each Task Group.
- A core-extensions approach should be applied in order to quickly bring
out a first consensus and prepare the next steps. Ideally the construction of
proposals for guidelines for
individual linguistic levels should be spiral-wise, in the sense of
making available a first draft as soon as possible, getting
feedback, amending and improving the proposal, circulating it again,
etc.
For the other levels, the work of the EAGLES lexicon group
should lead to a survey of the available approaches, as well as to a
methodology of comparing these proposals. Making such a methodology
available and testing it on a relevant subset of phenomena and
approaches is a goal in itself. A starting point may be work in the
EUROTRA-7 Study and in the NERC project.
The goals should be discussed by the participants to the task groups and
quickly decided for at the very beginning of the task work. Here follows a
proposal for goals for each task. This proposal is a general framework within
which each task will fix its own limits according to the input material
provided by a survey of available proposals.
- Architecture: core linguistic model, definition of the lexical levels and
their relations.
- Orthography: core of consensus features and comparison of specific
extensions (to be clarified later),
- Phonology: information needed by NLP applications,
- Morphosyntax: core of consensus features and comparison of specific
extensions,
- Syntax: core of consensus features and comparison of specific extensions
(to be clarified later),
- Semantics: aspects identified where steps towards consensus
building can most realistically be taken; directions for future work,
- Multilingualism: state of the art connected to the results of the other
levels (to be clarified later),
- Terminology: state of the art connected to the results of the other
levels (to be clarified later).
Next: Basis for comparison
Up: Methodology Task
Previous: Introduction