Modulo7 : A Full Stack Music Information Retrieval and Structured Querying Engine
Embargo until
Date
2016-01-18
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Johns Hopkins University
Abstract
In this thesis, the author proposes and implements a new Music Information Retrieval and Structured Querying Engine called Modulo7. Unlike other MIR software which primarily deal with low level audio features \cite{musicrecSurvey}, Modulo7 operates at a higher abstraction level, on the principles of music theory and a symbolic representation of music(by treating musical notes instead of acoustic pitches as the basic blocks of representation of musical data). Modulo7 is implemented as a full stack deployment, with server components that parse various sources of music data into its own efficient internal representation and a client component that allows consumers to query the system with SQL like queries which satisfies certain music theory criteria (and as a consequence Modulo7 has a custom relational algebra with its basic building blocks based on music theory), along with a traditional search model based on non trivial similarity metrics for symbolic music. Modulo7 also implements a lyrics analyzer, which supports functions such as lyrics similarity and meta data prediction (e.g genre prediction).
Description
Keywords
Music Information Retrieval