Newman et al. (2020)
Contents
Source Details
Newman et al. (2020) | |
Title: | Cover Song Identification - A Novel Stem-Based Approach to Improve Song-To-Song Similarity Measurements |
Author(s): | Newman, L., Shah, D., Vaughn, C., Javed, F. |
Year: | 2020 |
Citation: | Newman, L., Shah, D., Vaughn, C. and Javed, F. (2020) Cover Song Identification - A Novel Stem-Based Approach to Improve Song-To-Song Similarity Measurements. SMU Data Science Review, 3(2) |
Link(s): | Open Access |
Key Related Studies: | |
Discipline: | |
Linked by: |
About the Data | |
Data Description: | The study examines 80 source covers, each performed by two separate artists for a total of 160 songs. For each song, the similarity of specific audio features were analysed using a stemming process, resulting in four constituent components. TK Dataset: https://labrosa.ee.columbia.edu/projects/coversongs/covers80/ |
Data Type: | Primary and Secondary data |
Secondary Data Sources: | |
Data Collection Methods: | |
Data Analysis Methods: | |
Industry(ies): | |
Country(ies): | |
Cross Country Study?: | No |
Comparative Study?: | No |
Literature review?: | No |
Government or policy study?: | No |
Time Period(s) of Collection: | |
Funder(s): |
Abstract
“Music is incorporated into our daily lives whether intentional or unintentional. It evokes responses and behavior so much so there is an entire study dedicated to the psychology of music. Music creates the mood for dancing, exercising, creative thought or even relaxation. It is a powerful tool that can be used in various venues and through advertisements to influence and guide human reactions. Music is also often “borrowed” in the industry today. The practices of sampling and remixing music in the digital age have made cover song identification an active area of research. While most of this research is focused on search and recommendation systems, plagiarism is a real industry wide problem for artists today. Our research seeks to describe a framework of feature analysis to improve cross-similarity, song-to-song, similarity distance measurements. We do this with the context that cover songs represent a fertile training ground to prove methods that can later be applied to plagiarism use cases. Our proposed method preprocesses songs by first source separating the songs into its constituent tracks prior to feature generation. This is otherwise known as “stemming”. These subsequent spectral and distance features are then analyzed to provide evidence of improvement in overall modeling and detection performance. We find that using stem distances and overall distance measures achieves an uplift of 61.8% increase in Accuracy, a 59.2% increase in AUC, a 304.7% increase in Precision, and a 105.1% increase in F1 score for a regularized logistic regression. This process can be directly applied to more sophisticated deep learning frameworks”
Main Results of the Study
The study finds that whilst overall similarity (and vocalisations) of a cover song may be close when analysed using an algorithm, other more granular ‘stems’ are able to detect greater distance between perceived similarities (for example, in regards bass, drums, pitch etc.).
The methods explored in this article can potentially be used to detect song plagiarism. When analysing a sub-sample of the dataset of 46 songs, where plagiarism had previously been confirmed whether through court decisions or subsequent payment of damages, the final version of the stem model was able to confirm the same in 47.5% of cases. As such, this may be a step towards building an automated plagiarism detection system.
Policy Implications as Stated By Author
Whilst the study does not offer any explicit policy recommendations, the researchers advocate the inclusion of cover song detection algorithms to identify cases of plagiarism and infringement. However, they caution that this should be developed to aid individual artists, rather than large companies.
Coverage of Study
Datasets
Sample size: | 160 |
Level of aggregation: | Songs |
Period of material under study: |