Musimap’s cognitive technologies – revolutionising music recommendation
Here at the 4.5 Blog, we have two great loves: start-ups and music.
At Music 4.5, we have devoted two seminars – one in London, one in New York – to the power of the music playlist. In our discussion and debates about the processes and politics behind playlist creation, the subject of curation played a centre-stage role, always begging the question:
What’s better for playlist creation – human or algorithm?
Enter Musimap, which combines the two.
“Musimap is the European answer to The Echo Nest and Pandora’s Music Genome Project,” says CEO Vincent Favrat. “We are a cognitive technologies company that revolutionises music recommendation. We have developed a hybrid between human expertise and automated algorithm to make a humanized technology.”
How does Musimap’s cognitive technologies work?
Musimap emulates the complex emotional interaction that humans have with music. Its emotion-sensitive algorithm senses, reasons and responds to music as humans do. Thanks to the contribution of over 60 interdisciplinary experts (Daniel Levitin, Joseph Kerman, Gilles Peterson, Karl Bartos from Kraftwerk, etc.), Musimap has developed a cognitive computing technology rooted in a socio-psychological and musicological approach, as opposed to only relying on ICT-text retrieval or signal analysis.
“To decode music’s DNA,” says Favrat, “you need to take into account the whole palette of human emotions.”
Musip uses 11,322 keywords as well as 400 complex moods to describe music. “That goes far beyond traditional signal analysis, which proposes 5 or 10 moods top,” says Favrat. “One also needs to understand the mindset of each listener in its particular context., so we are using over 100 listening situations. To achieve a proper music recommendation, we believe one should also master emotions, relations and influences.”
Musimap’s granular database includes over 3 billion data points, 2 billion relations and soon up to 50 million songs. Its neural music network is the result of a unique blending between soft and hard science, where an in-depth human curator together with the latest AI technologies engineers a multi-layered system of crossed influences. “ In other words, we brought back soft criteria into the equation on top of hard science so as to achieve an emotion-sensitive music recommendation,” says Favrat.
“Unprecedented personalised recommendation accuracy”
He explains: “90% of the streams and sales comes from a mere 5% of the global music catalogue. The great promise of digitalization has only been half met. Yes, we have access to 50 million songs at one’s fingertips and that’s great, but listeners are now lost in an ocean of choices with no music compass to help them find their way in the overabundance of our hyper-connected era. Most music streaming platforms offer similar catalogues and services and their recommendations are limited by the dominating statistical approach or the sole use of human curation.
“This sounds like the past to me. With its humanized algorithm, Musimap unlocks music intelligence through cognitive technologies, where intelligence assistants can deliver highly personalized content recommendation in line with our IoT world and where technology recognises users’ intent and connects users with the right content at the right time. Our aim is to create the ultimate cognitive music assistant. That sounds like the future to me”.
Global validation of the technology
The accolades, awards and recognition of Musimap’s technologies are coming in fast and furious. Musimap was listed by Geektime.com amongst the top 10 companies in Belgium, was selected by Slush as one of the top 25 later-growth tech start-ups with the strongest growth potential, and SXSW labelled Musimap one of the 20 most innovative music companies worldwide, even using the Musimap API in the SXSW Music Hackathon. In addition, Musimap just welcomed renowned music tech expert Ted Cohen in our strategic board, always on the quest for the best available music recommendation.”
Scaling the technology to different market verticals
Musimap is already making plans for how its technology can extend to at least 20 different market verticals, including communication agencies, search engines, retail, connected cars, e-commerce, digital TV and radio, social media, digital assistant, and music streaming. At SXSW the technology’s relevance to other markets was in evidence when Cognitunes merged Amazon Alexa’s (Amazon Echo) voice-powered experiences, IBM Watson’s cognitive computersystem and Musimap’s emotion-sensitive technologies to create a prototype of future home digital assistant. “Our B2B positioning goes far beyond music streaming” says Favrat. “We’re targeting all businesses that use music as part of their offering.”
Ready to market
Musimap has closed a number of test collaborations with strategic partners and start-ups, including a music search engine for audio-visual professionals for Just Temptation, compilation and complex playlisting for Rendez-Vous Digital, and a distribution deal for the user experience in 4D amusement park rides for nWave Pictures.
Musimap is now focused on closing large-scale tests with clients in its key targeted segments. At Midem, Musimap announced the official start of a collaboration aimed at providing hi-res streaming service Qobuz with real-time personalised recommendations, a powerful search engine, and highly personalized radio based on the listener’s musicological profile. Musimap also just announced a partnership with European search engine Qwant to boost Qwant Music’s extensive search functions and enrich the user experience via advanced online tools.
Musimap is also about to finalize collaborations with international clients in the advertising and music sectors.
For now, Musimap is on the road, continually seeking collaborations and synergies by presenting at major industry events, including Viva Tech in Paris, the Montreux Jazz Festival, and the New York Venture Summit
“It’s a very exciting time for us,” says Favrat.