In order to understand politics, it is necessary to know about what political actors say saying and write. This understanding has special meaning in the Brazilian political system in which the organization of the Executive Power is based on great coalitions. In this sense, the present study makes use of Natural Language Processing as a tool for analyzing parliamentary discourse delivered in the Brazilian Chamber of Deputies and proposes a specific indicator, based on Chi-Square statistics, for the assessment of similarity of discourses. The results found are consistent with facts that marked the Brazilian political history in the period 2001-2015 and reveal that the ideological dimension is subject to the electoral logic in the formation of political alliances, suggesting that the proposed indicator has the potential to explain phenomena related to the behavior of coalitions parties.