1

As the number of textual analysis software programs increase, sociology is faced with the problem of gauging their suitability. On one hand, they save time insofar as they can process a large volume of texts but, on the other hand, the way in which the data is analyzed and the default hypotheses implemented in these tools require consideration. In this paper, we focus on the Alceste program and, based on a particular set of problems, attempt to deconstruct the way in which it processes text in general and semi-structured interviews, in particular. Reinert constructs a “summary” (among many possible alternatives) of the corpus considered as a whole by taking into account mainly hypotheses on words that concentrate the meaning of a text according to their grammatical category, but also hypotheses on text partitioning into Context Units, while seeking to remain within the “proper” conditions for applying certain statistical tools. This summary and analysis, based on regularities, do not, at any point, incorporate the sociologist’s hypotheses or theoretical viewpoint.