Searching digital music libraries David Bainbridge, Michael Dewsnip, and Ian Witten Department of Computer Science University of Waikato Hamilton New Zealand Abstract. There has been a recent explosion of interest in digital music libraries. In particular, interactive melody retrieval is a striking exam- ple of a search paradigm that differs radically from the standard full-text search. Many different techniques have been proposed for melody match- ing, but the area lacks standard databases that allow them to be com- pared on common grounds—and copyright issues have stymied attempts to develop such a corpus. This paper focuses on methods for evaluat- ing different symbolic music matching strategies, and describes a series of experiments that compare and contrast results obtained using three dominant paradigms. 1 Introduction There has been a recent explosion of interest in digital music libraries—indeed, Apple’s iPod has been called the world’s first consumer-oriented digital library. In all human societies, music is an expression of popular culture. Different gen- erations identify strongly with different musical styles. People’s taste in music reflects their personality. Teenagers, in particular, feel that their musical prefer- ences are strongly bound up with who they are. Many researchers seek to capi- talize on this natural interest by building digital music libraries [BD00,DB01]. Digital music libraries are an attractive area of study because they present interesting and challenging technical problems, solutions to which are likely to be highly valued by enthusiastic end-users. This paper addresses the problem of searching a music library for a known melody. In other words, given a fragment of an unknown melody, typically played or sung by a user, return a list of possible matches to a large digital library collection. This operation is not supported by the information structures provided by traditional libraries, except insofar as knowledgeable music librarians are able to provide human assistance. For scholarly work on melody, there is a book that provides a paper index of themes [Par75], but its scope is restricted to the older classical repertoire and it does not provide flexible searching options. And yet the problem of melody retrieval is of great interest to a wide range of potential users—so much so that there are popular radio programs that feature human abilities to “guess that tune”. A practical scheme for searching digital music libraries requires robust imple- mentations of several supporting components. First, it is necessary to assemble