Computational Processing of the Portuguese Language

In this paper we present a study about the rhetorical structure of opinion articles that have been written as part of college entrance examination. For that, we defined a set of rhetorical categories that aim at modeling the structure of opinion articles produced in this specific context and used it to manually annotate a corpus. Results of the annotation experiment showed substantial agreement among the annotators and disagreements were settled using the majority vote to build a gold standard corpus. This corpus was then used to build automatic classifiers that assign one of the possible categories to each sentence of the opinion article. Experimental results regarding the classification were promising considering the model’s simplicity and the reduced number of training instances.