Alexandru Timotin, Florin Teodor Tãnãsescu *
Structures for a Thesaurus of Technical Terminology
Each concept-except the one designed
by the title of the first domain-has at least an ascendant relation
and rarely more than four. As already mentioned, one and only
one of the ascendant relations has to be chosen as the concept
priority relation, which allows its insertion in the taxonomic
tree-like structure of the thesaurus.
While a multilingual thesaurus
based on a standardized terminology has been developed, the introducing
of the semantic relations of a concept tries to suggest the source
definition (if any) or the meaning of the source context (for
not explicitly defined concepts) by using all the available source
languages (three in the case of the IEC thesaurus). Our experience
shows some specific aspects:
-
The definitions are often too particular
(i.e. conceived merely for the narrow application area of the
chosen source publication), they are different from one source
to another (if more sources are available) and they sometimes
have different meanings in the different languages of the same
source. These difficulties are less apparent in the last edition
of the IEV, which is a highly structured vocabulary, but are inherent
when many years separate the various successive chapters, for
the technical progress is too fast. In consequence, a thesaurus
has to integrate the source information on a larger time interval,
and has to give a less restrictive semantic vicinity than
a definition does.
-
The selection of the ascendant relations
has to follow a lot of rules, but cannot be done according to
a precise algorithm. This selection is essentially pragmatic and
guided by the interest to obtain a suggestive result with a lesser
number of ascendant relations. It is rarely unique and will change
according to the development of the available concept fund.
-
The number of necessary types of basic hierarchical relations
is much greater than five. For instance, all the special cases
of applications of the vague RL relation (see §3.2.1.5)
are distinct and very precise relations and might have been treated
as independent types. The choice of only five basic relations
is pragmatic and, therefore, not unique.
-
Not all concepts
have to be treated similarly. A minority of high-level concepts,
the "zones" - top concepts of the chosen semantic fields
- and some of their 1...3-degree descendants, are important and
their ascendant relations have to suggest their definitions. Such
concepts also have a rich semantic first-order descending, that
completes in a very efficient manner their semantic vicinity (Table
2).
-
The semantic vicinities may be verified by considering tree-like
lists of descendant concepts on many levels (see §4.3.2).
Such lists have to exclude "vicious circles" obtained
from a chain of relations of the same type (homogeneous
chain) and have to avoid such circular heterogeneous chains (obtained
with different types of relations). The presence of DO
relations can generate such chains, logically correct: the mechanical
work is a form of energy transfer but the energy is defined by
using the mechanical work as an originating concept.
4. Proposed lexical structures
4.1. Term category
The lexical structure of the thesaurus, language-dependent, is
defined at first by relating each term of that language to its
corresponding concept, language- independent. This term-to-concept
relation is expressed by a term property called term category,
which indicates its "level of use" in this language:
Term category
| Symbol
| Graphic reference
|
1. Descriptor (recommended term)
| @
| .
|
2. Accepted synonym
| T
| => descriptor
|
3. Not recommended, obsolete term
| N
| != descriptor
|
4. Recommended abbreviation
| A
| .= descriptor
|
5. Recommended literal symbol
| L
| L= descriptor
|
6. Alternate, shortened or permuted, form*
| S
| -> descriptor
|
* With a unilateral reference to the descriptor.
113