Our work aims at the optimization of existing tools for computer-assisted description and analysis of textual data. More specifically, we have been involved in the thematic description of clauses and clause complexes of Quebec budget speeches from 1934 to 1960. Our main objective is to enhance the work already done in this direction by elaborating the analytic framework through a study of the thematic structure of these discourses. We first set out the general context of our work by briefly explaining the research project on political discourse under the Duplessis Regime in Quebec (1936–60) and giving a brief survey of the parsing strategy applied to the corpus. Second, we present the theoretical background of thematic analysis and the operational model that we are using here. Finally, we try to illustrate the relevance of such methodological work on research data.