Please use this identifier to cite or link to this item:
|Title:||Kleptamatic: A System for Determining Musical Similarity Via Discovery, Clustering, and Classification|
|Abstract:||Music’s effect on humans is undeniable. It can inspire, comfort, enrage, depress, and ultimately tell us that regardless of how we may feel, we are never alone. Because every piece of music makes different connections with humans, what people choose to listen to is what is most suitable to their temporal feelings at the moment. Liking an artist or track can involve many external or undocumented factors such as historical context, personal history, lyricism, and peer enforcement. Because of this, a reproducible and predictable model to guide the choice of music for listeners is too complicated to develop and measure its accuracy definitively. However, technical advances make it possible to analyze the phenomenon of this additive synthesis of sine waves that reaches people on an emotional level, akin to Nirvana. Current music recommendation frameworks seek to suggest music based on a combination of qualitative and quantitative features present within a musical signal: if a listener likes a song or artist, he or she will likely like similar songs or artists based on some shared characteristic in the signal or the artist’s inspiration. This thesis aims to develop a computational model that improves the process of musical discovery beyond existing ones. This new model defines and ascribes the notion of musical similarity using a combination of quantitative and qualitative features of music. Songs and artists always have some measures of similarity, although they are not easily quantifiable or tangible. Listeners have built these associations within music in terms of genre labels, social tags, and playlists from their personal experiences with music. The goal of the proposed recommendation model is to accomplish a more expanded analysis of factors involved between music and listeners. It has three principle components: (1) Arachne, a web crawler to collect the names of trending tracks and artists, both new and established, (2) Trieur, a module to analyze collected songs, cluster by acoustic features, and classify by high level labels, and (3) Kleptamatic, an interface to explore the musical space based on the clustering and classification of different musical samples. Analytic integration of each component guides listeners to choose the right music for the right moments.|
|Type of Material:||Princeton University Senior Theses|
|Appears in Collections:||Electrical Engineering, 1932-2017|
Files in This Item:
|PUTheses2015-Park_Hannah.pdf||2.45 MB||Adobe PDF||Request a copy|
Items in Dataspace are protected by copyright, with all rights reserved, unless otherwise indicated.