Movie stills of the Durcheinander Presentation at Animax, Bonn in 2006.

Movie stills of the Durcheinander Presentation at Animax, Bonn in 2006.

2008 by [intlink id=”2″ type=”page”]Till Bovermann[/intlink], and Julian Rohrhuber.

With Durcheinander we present a system to help understand Agglomerative Clustering processes as they are used in various, often visual oriented, data mining and exploration systems. Durcheinander consists of several small objects on a tabletop surface, which represent data items in an artificially generated data set. A computer vision system tracks their position and computes a cluster dendrogram, which is sonified every time a substantial change in this dendrogram takes place. Durcheinander may be used to answer questions concerning the behaviour of clustering algorithms under various conditions. We propose its usage as a didactical and explorative platform for single- and multi-user operation.
Agglomerative Clustering is a data mining approach that is mainly used to unveil structural relations in high-dimensional data sets. It particularly facilitates the discovery of compact clusters of data items in high-dimensional vector spaces. Structures found by Agglomerative Clustering are assembled into a dendrogram that recursively interconnects single data items by means of their location. Although the general behaviour of Agglomerative Clustering with a given set of meta-parameters (which includes the used distance metric) can be easily understood, the parameters’ relation to the algorithm’s result in a specific case is more difficult to grasp. Participants in data mining courses can achieve better understanding by trying to answer the following questions:

  1. Under what variations in the data does the Agglomerative Clustering dendrogram
  2. What happens when data items are in a special configuration?
  3. What are the differences between the various distance metrics?
  4. What are the differences between the various cluster metrics?

Durcheinander’s purpose is to help answer these questions by means of the TAI paradigm. It provides the opportunity to physically grasp the data and, at the same time, allows auditory exploration of the effect of different clustering parameters. Durcheinander’s tangible objects are laid out on a table and the sound is delivered in a spatial sound environment. Learners have turned out to particularly benefit from this collaborative multiuser nature of the system; it invites to discuss the results of Agglomerative Clustering in the process of co-operative exploration, instead of before and after. Furthermore, its interactive programming approach allows researchers to experiment with different Sonification methods during interaction.

Example for a hierarchical clustering.

Example for a hierarchical clustering.

Additional Material

  • Original Publication (pdf)
  • PhD Thesis pp. 162–166 ([intlink id=”102″ type=”page”]Publications page with link to PhD Thesis[/intlink])
  • Video of the presentation in 2008 at Animax, Bonn.

People involved in the Production Process

Till Bovermann, Julian Rohrhuber, Thomas Hermann, Helge Ritter.



	Address = {Paris, France},
	Author = {Bovermann, T. and Rohrhuber, J. and Ritter, H.},
	Booktitle = {Proceedings of the 14th International Conference on Auditory Display},
	Month = {June 24 - 27},
	Year = {2008}}