Analyse discriminante et vocabulaire politique : Essai méthodologique

Authors: , ,


This paper aims at testing the hypothesis of growing ideological uniformity of political speeches. If political speeches lack ideological differences, it should be difficult to re-classify them only by analyzing the presence or absence of lexical items. We first worked out a method to classify political speeches and then carried a test on two speeches by leading Belgian French-speaking politicians.  The method is based on discriminant analysis. It utilizes the words most encountered in one speech and not in the other as discriminant factors. Statistical softwares then assess a discriminant function used to re-classify short parts of each speech called blocks. The most discriminating 10 factors re-classify correctly 89% of the blocks. The percentage increases to 93% with 20 factors and to 98% with 30 factors.
However the results should be taken with caution because of the limited sample, the test tends to question the growing uniformity of political speeches. The sampled ones had enough specific features for allowing a rather simpte method to re-classify most parts of them correctly, even if some typically ideological items are not to be found.


How to Cite: Chapelle, J. , Couvreur, P. & Pagano, G. (1992) “Analyse discriminante et vocabulaire politique : Essai méthodologique”, Res Publica. 34(1). doi: