This article describes ClusterXplain, an approach helping users to better understand the results of their queries. These results are structured with a clustering algorithm and described using a personal vocabulary. We present a crisp and a fuzzy version of this approach. The goal is to find what the elements of a cluster have in common that also differentiates them from the elements of the other clusters. The data considered for characterizing each cluster of answers are not limited to attributes used in the query, revealing unexpected correlations to the user. We provide users with characterizations using terms from the natural language to describe the obtained clusters.