Le Logiciel Prospero A L'Epreuve D'Un Corpus De Resumes Sociologiques

Abstract

The Program Prospéro Tried Out Against A Corpus of Sociological Abstracts: The objective of this article is to test the potential of the program, Prospéro, for analyzing the abstracts of the first congress of the French Sociological Association (AFS). Prospéro was conceived as a program for the analysis of corpora of texts characterized by a certain variability (particularly in the use of argumentation and time). We consider these abstracts as a “dossier; that is, as a collection of texts characterized by the unknowns and the obligations of authors proposing a presentation in a Thematic Network (TN) session. To begin with, we show how much receding – done by the researcher in the case of Prospéro – is decisive in determining the level of analysis and of hypothesis testing. We then propose a model describing the expectations of a potential presenter and construct categories, collections and fictitious entities to test this model. The program, in this context, functions as an instrument for identifying singular properties of the corpus (for example, judging the usefullness of a category or a collection in characterizing a TN). But it also identifies atypical texts which can encourage the modification of the model. Finally, we study how presenters adapt to the requirements of a proximity between the abstract and the call for papers distributed by the TNs. This approach shows that only a few TNs bring together texts sharing similar characteristics. This work permits us to identify certain interesting aspects of French sociology and, in particular, the possibilities and the limits of Prospéro. The program functions as a research tool which can test a researcher's interpretations, hypotheses and models. The program would be even more useful if corpora could be complemented by texts permitting a variation (particularly historic) which would benefit the analysis of sociological scientific production.