The Quantitative Evaluation of the Pathos to Ethos Ratio in Scientific Texts

In this paper we present the annotation-based argumentation analysis of scientific texts with the aim of quantitatively evaluating their persuasiveness in terms of two infrequently studied persuasion modes: ethos and pathos. The expert annotation of texts is performed manually through the use of web-tools for visualizing argumentative statements and reasoning schemes, along with representing the argumentation structure of a text in form of a directed graph. In these graphs, argumentative statements and connecting arguments (reasoning models) are correspondingly denoted by information nodes and scheme nodes. We describe a method to evaluate the argumentation in a text for ethos and pathos distinctly, by analyzing both the content of information nodes (premises and conclusions) and the functionality of corresponding reasoning models. For evaluating information nodes, we have constructed two dictionaries of argumentation markers that are represented by lexical-syntactical patterns and respectively identify ethos and pathos in collection texts. Quantitative values are calculated for 30 annotated texts of the collection. We employ these values to group these texts into several sets based on the proportion of pathos and ethos.