Over the last three years, my colleagues and I at Josplay have been working on something unique and at the foundation of human existence — the intersection of music and emotions.
The driving question for us was: how can we use African music to enrich and satisfy the emotional needs of Africans while respecting the nuances in our group listening behavior?
It is a universal knowledge that music is the food of the soul. It is the most common emotional pill for most humans. We turn to chants and hymns when we want to meditate or be solemn. We use up-tempo-high-valence songs when we want to celebrate or be joyful. When we just want something in the background to help us focus on other tasks, we go for songs that do not draw our attention to any extremes of the emotion spectrum.
The journey has been exceptionally challenging and equally rewarding.
African music is rich and complex. In most cases, it is a cocktail of rhythms and innovations that draw various kinds of emotions from different listeners.
What might sound pleasant to a listener in Enugu, Southeast Nigeria, may come off on the opposite end of the emotional spectrum for a listener in neighboring Markurdi, the capital city of Benue state, North Central Nigeria.
We listen along cultural lines and contexts. Therefore, music can be sacred to us.
The core challenge for us is: how do you organize and codify these peculiarities such that intelligent software applications can accurately match music to listeners at both the cultural and emotional levels?
We realized, early in the project, that there is no publicly available knowledge and metadata repository to aid our R&D.
Metadata, the fact about a piece of music, is different from the music. It answers The Who, Where, and What about the music. It is critical for various purposes — including display, utilization (playlisting and recommendations), and royalty distribution. Yet, it is the most neglected in the music-making process.
Currently, where these metadata exist — especially for recent releases — they are held in silos across the music industry chain. Older works hold these data on the covers of vinyl, cassettes, and CDs.
Collecting and analyzing these data is a painstaking task that involves human musicologists and machine learning models. Generating fresh ground truth data to power our machine learning processes has been equally arduous.
I will share more about our methods in subsequent updates. From what we’ve learned, it is not an NP problem. It is tractable but not trivial.
Today, we are sharing some raw artifacts that have helped in this journey through the African Music Library. The Library is a repository of both data and knowledge. Data about who did what and when. In the Library, you can browse an artist and drill down to each track in their discography to get an audio analysis. You can find people and businesses who participated in making the discography.
The Library also documents how the music is made. Visitors can find knowledge on several music genres on the continent as well as compare genres. And, yes, not all music from Africa is Afrobeats. Neither are they the infamous ‘world music’.
As time go by, we will iterate and add more features to satisfactorily realize the Library’s goals of being the most comprehensive repository of music knowledge and metadata on African music.
Our short-term goal is to scale the volume of data available for public use. We will also publish our APIs to give other application developers access to our data.
We hope the Library will galvanize a community of musicologists, creators, and engineers. And ultimately, we hope it stimulates the much-needed knowledge and tech innovation in the African music space.
We hope it will be a launching pad for the great ideas that will follow and that innovators who come afterward will have the data they need to go further than we can ever reach.
Founder, African Music Library
Thanks for reading African Music Library Newsletter! Subscribe for free to receive more updates on the African Music Library Project.