Analyse lexicométrique et recherche des cooccurrences

LEXICOMETRICS AND COOCCURRENCE SEARCHES In this article, P.L. seeks to identify the fairly close relations within a text string. The material criterion used to study these relations is the cooccurrence, that is, an observation of recurrent proximity between the same text- form. P.L. examines three possible methods which would enable us to measure the phenomenon of cooccurrence and hence proceed to the automatic selection of the most closely related forms. In order to identify cooccurrence, the idea of a pole on which the first method is based is seen in a critical perspective and is replaced in the two other methods by the notion of pair and couple. On the basis of some experiments carried out on a text, P.L. suggests, however, that the second method is not entirely appropriate to analyse the problem in hand. The third method based on the measure of the number of couples and pairs found in the same sentence is satisfactory. This last method has been described in entirety and tested on the general resolution of the CGT congress in 1975.