Interpretative data analysis: when semantics makes deflect the number. The case of large scale textual corpora.
Résumé
The mathematical foundations, methods or models of an approach, are not a guarantee of objectivity, and the lack of a sustained reflection on an authentic interpretative data analysis (IDA) increases the parts of relativity of what initially has been thought as objective. The aim of this paper is to give some argument about the urge and the importance of a reliable hermeneutical approach in data analysis. In the first part, we discuss the epistemological necessity of such an approach. In the second, we argue on the feasibility conditions of IDA and try to set up an architecture in which humans and machines may share competences and performances in interpretative targets. We specially focus our attention on the interaction protocol between human and machine and reveal the crucial role played by the notion of isotopy. In the third part, we outline a system that fully implements the basic principles of IDA. We finally apply such a system in the case of content driven data analysis of textual corpora.