Policy Distance Determination Methodology Based on Topic Analysis
Digital Humanities, Topic Analysis, Programmatic Dimensions, Political Distance, Metrics
This study aims to contribute in the field of Digital Humanities with an automated analysis of the performance of political parties, presenting a new approach that includes an ideological aspect in the treatment of the multiple programmatic dimensions expressed in manifests that depict the role of these associations in society. The proposed approach avoids syntactic analysis and prior annotation of documents and is based on statistical inferences in a lexical analysis of the content of manifestos. Like other approaches found in the literature, the proposed procedure consists of carrying out empirical measures of distance on ideological polarizations or distinguishing programmatic issues, necessary to test spatial models of political action. The focus is on the use of natural language processing techniques to extract the multiple dimensions - of party performance and the theoretical line under which the analysis will be carried out - from a corpus of theoretical works for later comparison, in the light of theory, of the political distance or degree of affinity between the associations.
The thematic sequence of the dissertation consists of the presentation of different approaches to content analysis of party documents and government coalitions, the metrics used for this purpose and the applications in Brazil. Finally, we present a new methodology based on topic analysis and the preliminary result achieved.