THEMATIC WORDS AND PHRASES

Many definitions have been proposed for thematic analysis (Louwerse and van Peer, 2002). According to Popping (2000), thematic analysis can be described as the “identification of what, and how frequently, concepts occur in texts”. This definition, like many others, remains very general and fuzzy. Stone (1997) mentioned that the concept of theme “is used in a loose, general way for analysing patterns in text”. Stone’s point of view remains very general, but seems to allow some type of computational perspective on theme analysis. However, instead of describing theme analysis in terms of patterns analysis, the literature on this topic (here, we refer to studies in philosophy, literature, text analysis, etc.) shows that theme analysis lies more in the discovery and identification of the multiples relations between different themes that make a textual corpus consistent and intelligible. We will show how this definition can lead to some type of computer-assisted theme analysis.

In general terms, a classifier is an abstract machine or function that realizes some type of grouping or sorting on a set of objects. In a set theoretical definition, classification is the projection of a partition on objects so that they are as homogenous as possible. In categorical terms, a classifier can be defined as a quadruplet (O, X, I, G) where:

O is the set of objects (o₁…. o_n);

X is the set (x₁…. x_n) of features describing each object O;

I is the set of types (i₁…. i_n);

G is a discriminant function.

In this perspective, a classifier is conceived as a categorical operation defined in the following manner:

For each object O((G(x₁…. x_n) i)x_j).