A Method Based on the Ontology of Language to Support Clusters’ Interpretation

Abstract

The clusters' analysis process comprises two broad activities: generation of a clusters set and extracting meaning from these clusters. The first one refers to the application of algorithms to estimate high density areas separated by lower density areas from the observed space. In the second one the analyst goes inside the clusters trying to figure out some sense from them. The whole activity requires previous knowledge and a considerable burden of subjectivity. In previous works, some alternatives were proposed to take into account the background knowledge when creating the clusters. However, the subjectivity of the interpretation activity continues to be a challenge. Beyond soundness domain knowledge from specialists, a consensual interpretation depends on conversational competences for which no support has been provided. We propose a method for cluster interpretation based on the categories existing in the Ontology of Language, aiming to reduce the gap between a cluster configuration and the effective extraction of meaning from them.