OLINDDA (OnLIne Novelty and Drift Detection Algorithm) addresses the problem of novelty detection in an online continuous learning scenario as an extension to a single-class classification problem. This paper presents its current version, that evolved toward the discovery of new concepts initially as emerging clusters and further as cohesive sets of clusters. New strategies for validation and merging of clusters as well as for dynamically adapting the number of clusters are discussed and experimentally evaluated.