Expertise and strategies in scientific documents indexing. This paper
presents the results of a study of scientific documents indexing. Inde
xing consists in identifying the main ideas of a document and in conve
rting them into indexing terms chosen among a controlled vocabulary. I
n the area of information sciences, several studies have shown an impo
rtant variability among indexers. The characteristics shared by the in
dexing and diagnostic tasks enable us to consider various indexing str
ategies according to expertises, these strategies allowing us to expla
in the variability among the indexing terms chosen. Twenty-five subjec
ts have indexed eight books; they had to write the terms depicting the
main ideas of documents using the indexing terms they extracted from
the documentary language RAMEAU. Two categories of variables have been
controlled: the textual content (two scientific fields: economics and
psychology) and the operators' characteristics (expertise in indexing
and in use of the documentary language RAMEAU and familiarity with th
e scientific domain). The dependent variables are the number of indexi
ng terms chosen, the rate of the intra-indexer consistency, the nature
of indexing procedures, organisation of these procedures and, finally
, the role of the level of expertise on the indexing strategies used.
Results indicate three indexing strategies: ''indexing driven by the k
nowledge of potential indexing terms'' ''indexing aided by the documen
tary language'' ''indexing pointed at a user profile''. The first one
was used by the professional indexers who practice the documentary lan
guage RAMEAU; it's characterized by a search for indexing terms from t
he document analysis stage. This search appears through a high rate of
the intra-indexer consistency and through the objectives pursued. The
se indexers carried out a level of analysis corresponding to the level
required by the language RAMEAU. The second strategy was used by the
beginner indexers; it consists in identifying the generic terms in the
document that allow them to access quickly to the documentary languag
e used then like a framework search. The use of this strategy goes wit
h the lowest rate of similarity between concepts and indexing terms. T
he third strategy was used by the professional indexers working with a
nother documentary language. Their analysis was deeper; the subjects c
hose a more important number of indexing terms and felt some difficult
ies to convert concepts into indexing terms. These results raise quest
ions about the compatibility between knowledge representation produce
by the documentary languages and the operators' representation in orde
r to elaborate new documentary language or to improve existing languag
es.