This work presents a methodology that proposes the combination of quantitative and distant-read data with detailed, closer reading in discourse analysis, enabling new possible views over data and diverse perspectives of analysis. This methodology makes use of resources most used in corpus-based linguistic, such as frequency lists, preferences, categorization, and reading concordance lines. Its application is demonstrated using as exploration object Brazilian literature titles in the public domain, compiled in a corpus with approximately 5 million words, semantically and morpho-syntactically annotated, and by using computational tools that enable searches based on lexical-syntactic patterns of the Portuguese language. The purpose is to identify how the male and female characters are portrayed in those texts, enabling the creation of a general view on how women and men are built through language. The study happens in two fronts: by observing the predicates used on describing characters and the actions these characters take, comparing the male and female results and analyzing them in a critical way.