Une application de la fouille de textes : l'extraction des règles d'association à partir d'un corpus spécialisé
Jérôme Azé et Mathieu Roche
Abstract
In many domains (biology, medicine, psychology, etc.), efficient text mining tools could help the expert.
In order to obtain a usable tool, an expert of the domain must control the various text mining steps.
The approach proposed in this paper consists in extracting the association rules specific to the field starting from a set of specialized and homogeneous texts.
Our approach is made up of various stages in which the expert's role is essential.
The first stage consists in extracting the terms in the texts and associating them to a concept, i.e. a set of terms having the same semantic.
Using this new specific knowledge, the initial corpus is transformed into a matrix.
At the last stage of our approach, this matrix is discretised in order to extract association rules.
Keywords
Taxonomy, Terminology, Discretisation, Association Rules.
