What was the customer’s Business Need/Challenge?
Client is a commercial music streaming service providing free and licensed music and provides both Indian and international music content. To build a recommendation engine to suggest relevant songs to users based on their pattern of listening and choice of music. Client has a warehouse of more than 10 millions untagged songs, so we had to build an automation system to tag these songs by Genre, Male/Female Voice, Artist, Style etc.
What solutions has Fragma delivered to drive the consumption of Azure?
We used advanced music analyzing libraries in ‘R’ to extract features from wave/mp3 files. We splitted the song in to ‘5sec’ intervals batches, and extracted features for each batch. For each batch, we extracted features like MPCC, Pause rate, Pitch, Timbre, Acoustics etc upto 60 features. Then used clustering to segment these songs, the segments were tagged with Genre Happy, Sad, Romantic etc
What impact did Fragma drive, how did the solution/offering address the customers business needs?
Helped client to automate tagging of Millions of songs and segment similar songs. Recommendation engine was built on top of segmented songs for personalized customer experience.
The songs dump was stored on blob, R was used to extract upto 60 feature and were then stored in Azure SQL. Based on user profiles and Genre of songs, a recommendation prototype was built using Cognitive Services Recommendation Engine.