A Semantically-Based Computational Approach to the Annotation of Narrative Style

This work describes the annotation of the novel "The Solid Mandala" (Patrick White, 1966), carried out combining sentiment and opinion mining on character level with the Appraisal Theory framework, here used to identify evaluative statements and their contribution to the social dimension of the text. Our approach was inspired by research on the correlation between White’s style and the personality of his main characters. The annotation was manually executed by second author using an XML standard markup system and double-checked by first author. In this paper we comment on the selected features, focusing on the ones acquiring specialized meaning in the context of the novel, and provide results in terms of quantitative data. Comparing them, we are able to extract story units in which special or significant events take place, and to predict the presence or similar units in the narrative by detecting concentrations of features. Eventually collecting all annotations has made available a lexicon where all ambiguities are clearly identifiable and verifiable. The lexicon will be used in the future for the automatic annotation of other novels.