A tech disruptor in music and entertainment, wanted to create a product that used the emotional content of music to deliver personalised listening experiences. Early prototypes, built using mainstream data science methods, failed to capture accurate emotional signals across a diverse range of music. Optima applied GenAI methods to extract structured insight from raw, unlabelled audio data, improving accuracy and enabling a scalable solution.
Context:
The company set out to build a product that used the emotional content of music to deliver hyper-personalised listening experience. The main obstacles were the lack of labelled data and the limitations of conventional methods, which made it difficult to extract reliable emotional signals at scale. Early prototypes struggled to achieve the accuracy required, putting the product’s differentiation in a competitive market at risk.
What we delivered:
- A robust and reusable framework to turn raw, unlabelled audio into structured business insights.
- A highly advanced GenAI system capable of extracting latent features that captured the emotional make-up of music.
- Automated, scalable end-to-end pipelines that ingested audio files and generated unique emotional signatures across thousands of tracks.
- Outputs that were interpretable and aligned to the client’s product needs in categorisation and personalised recommendation.
Outcome:
- Tangible business insights derived from unstructured, unlabelled audio data.
- Significant improvements in the predictive accuracy of emotion models.
- A product development pathway that was robust and reusable, future-proofing the organisations offering in a fast-moving industry.
- Stronger ability to serve personalised listening experiences to a diverse and evolving customer base.
‘’By applying state-of-the-art AI, we built a product that turned raw audio signals into a unique emotion digital fingerprint for any piece of music. The product provided superior accuracy of emotion recognition and gave the client the foundation to deliver hyper-personalised music experiences at scale.’’
Dan Blagojevic PhD – Chief Data Scientist
