Select a clustering scheme:

This tool is dseigned to help surface songs from the mashup corpus for the various machine clustering attempts we've made at organizing the data, to facilitate qualitative evaluation of the clustering schemes alongside the extant quantitative evaluation.

To use this tool, select a clustering scheme from the drop-down on the left. The page will update with example mashup/source songs from the clusters generated by that scheme, as well as visualizations of that clustering scheme's general data trends. The links at left to the entire corpus will also update to reflect the cluster placement of each track.

To start with clustering schemes that are likely more robust, you might consult this table. This shows a quantitative grading of the sensitivity of each clustering to a single variable and the diffuseness of the clusters. Clusterings that are very sensitive are less robust as they are likely only measuring the single variable to which they are sensitive. Clusterings that are very diffuse may have less concise and clear categorizations. I would prioritize avoiding very sensitive clusters, and within what is left, favor the less diffuse ones.