This document proposes a method for clustering and labeling clusters using a double application of self-organizing maps (SOM). The method uses tensors to represent multi-aspect data and applies SOM first to cluster the data nonlinearly. It then applies SOM again to the cluster representation to visualize the main concepts within each cluster, which can then be used as potential cluster labels. The method is demonstrated on a case study of finding thematic research areas within a future cities laboratory by clustering researchers based on their interests and applying the double SOM to identify and label the main research clusters.