Ordinal-Contextual Dissimilarity for Analysis of Heros in Tragedies

Bahman Afsari, Katayun Mazdapour and Bruno Jedynak
Proceedings of Bridges 2015: Mathematics, Music, Art, Architecture, Culture (2015)
Pages 301–308 Regular Papers


One of the goals of a literary analysis of the tragedies is to identify the relationships among the heroes. We aim to provide computational tools which can help in this task. Ideally, this analysis would be based on the raw text. Instead, we rely on tables extracted by researchers from the text which summarize objectively the qualities of each hero from three famous classic stories in the available sources: Trojan War from the Homer's Iliad, Julius Caesar from the Shakespeare's play, and knights in the Ferdowsi's Shahnameh, a Persian epic story. We develop a statistical analysis for identifying clusters (groups) among a group of characters in a story starting from such tables with ordinal data. We define the Ordinal-Contextual (OC) similarity (or equivalently the OC dissimilarity) as a linguistically meaningful and context-based similarity (or dissimilarity) for ordinal data. Using classical multidimensional scaling, we represent the heroes in 2-D space and cluster the heroes using the affinity propagation technique. Overall, this analysis turns out to confirm and add quantitative validity to the hypothesis that the rival heroes in these classic tragedies who constitute the core of the tragedy are similar and distinguishable from the other heroes.