This paper presents TuneGraph, an online visual tool for exploring melodic similarity. The underlying data comes from a large index of online music, all transcribed in abc notation, and TuneGraph uses a melodic similarity metric to derive a proximity graph representing similarities within the index. A rich but dense graph is built and then sparsified by removing weak, non-essential edges. From this a local graph is extracted for each vertex, aimed at indicating close variants of, and similar melodies to, the underlying tune represented by the vertex. Finally an interactive user interface displays each local graph on that tune's webpage, allowing the user to explore melodically similar tunes.
Keywords: cultural informatics; music similarity; force-directed placement; search visualisation