Visual analysis of interactive document clustering streams

Abstract

Interactive clustering techniques play a key role by putting the user in the clustering loop, allowing her to interact with document group abstractions instead of full-length documents. It allows users to focus on corpus exploration as an incremental task. To explore Information Discovery's incremental aspect, this article proposes a visual component to depict clustering membership changes throughout a clustering iteration loop in both static and dynamic data sets. The visual component is evaluated with an expert user and with an experiment with data streams.