Tune-Finder: An Application for Discovering Similar Melodic Motifs

Abstract

In this paper, I describe the process of building and evaluating Tune-Finder, a tool which provides a means of humming a tune in order to find other tunes (primarily folk and classical compositions) with similar melodic motifs. Here, the term motif is used interchangeably with the term contour to signify a sequence of characters which represent the melodic relationships between consecutive pitches in a tune. The tool is a NodeJS web application built using the Express framework and deployed using AWS Elastic Beanstalk. There are three primary components which comprise the interface: a visualizer which displays audio input, a search bar providing the ability to type input representing a motif, and a results page which displays the most similar tunes to the user’s query from a corpus. Furthermore, the tool allows users to select a tune from the results page and assess their accuracy when humming the chosen tune. These features together allow music-loving individuals to uncover some of the origins for popular contemporary motifs and perhaps find music that they may be interested in. Besides facilitating the discovery of music, Tune-Finder is designed to be an interactive tool that anyone can use as a means to test out and improve their sight reading, pitch matching, and humming abilities.