The present study offers a novel methodology for corpus-based discourse analysis that combines Kenneth Burke's (1968, 1969) dramatistic method of text analysis and the corpus techniques of keyword extraction and concordance reading. Applying the methodology, a two-stage analysis of Donald Trump's 2016 Orlando speech has been conducted: First, at a micro level, (i) the keywords used by Trump were identified in his speech as compared against the historical reference corpus of US Presidents through WordSmith Tools (Scott, 2012), and (ii) Trump's keyword lexical structures were described and interpreted in the pentadic ratios rhetorically motivated by the "terministic screens" that select and deflect representations of the gay-nightclub-shooting event in the speech; second, at a macro level, the process of substantiating the whole speech event of Orlando was implemented in terms of the different types of substance -- geometric, familial, and directional -- recognized in the keywords clustering and linking in the speech. The study has reached two findings. First, methodologically, Burke's dramatistic method of text analysis and corpus-linguistics techniques have been synergized in a novel way that contributes to analysing the rhetorical language of political discourse, eventually yielding a corpus-based dramatistic discourse analysis. Second, practically, analysing the pentadic ratios of the speech keywords and their substance types has revealed Trump's rhetoric of selecting and deflecting certain representations of the Orlando political event through a specific ensemble of terministic screens and substantiations.