How to Extract Knowledge from Texts. Data-mining and Topic Modeling Applied to a Dantean Collection

Abstract

This paper intends to illustrate an academic methodology applied to a digital Dante repertory – the Database Allegorico Dantesco (DAD) – developed within the framework of my doctoral research and containing hundreds of texts related to the allegorism of the Divine Comedy. By addressing specific computer theories to the archived works and by resorting to the joint use of several digital instruments, a study procedure will be described that, by referring to statistical principles of applied sciences, will allow for objective results in relation to the theoretical discussions developed around Dante’s allegory in the 20th-21st centuries. Specifically, it will be shown how it is possible to subject the corpus of selected texts, since they have already been transcribed in XML for archiving in the DAD, to topic modelling in order to identify groups of semantemes referable to authors, motifs and recurring themes in the allegorical readings of the poem to be used as guidelines for a direct and in-depth examination (close reading) of the object of the research. In this way the digital tools will be actively used for the automated extraction of different types of data (data mining) corresponding to thematic nuclei to be developed in order to trace a critical profile of the allegorism of the Comedy.