motiv

Inspiration

We wanted to change the way people experience music.

What it does

We created a tool that categorises music based on emotion.

How we built it

We used Google Cloud Natural Language API to attach a sentiment value to each song's lyrics. We combined these values with values attained from our Recurrent Neural Network, built with Keras and with hyper-parameters optimised by Bayesian Optimisation. The input space contained mp3 files that we analysed and created a feature space for which was then inputed into the RNN as a 3D Tensor. We also used a natural language toolkit to cluster data into categories that allowed us to find common words used in a specific emotional category. We combined the 3 results to calculate a final category for each specific song.

Challenges we ran into

Almost close to everything.

Accomplishments that we're proud of

Everything.

What we learned

Everything.

What's next for motiv

We want to improve this product so that we can target specific markets like museums, musicians and marketing.

Built With

• bootstrap
• css
• google-app-engine
• google-cloud
• google-cloud-natural-language-api
• google.
• html5
• intel
• javascript
• json
• keras
• machine-learning
• matplotlib
• nltk
• numpy
• pandas
• python
• recurrent-neural-network
• sklearn
• tensorflow