Mandarin is a language that is know for it's difficulty to learn. Mandarin has four tones which can make non-native speakers have difficulty learning the language.

What it does

This web app displays tonal words in pinyin so user's can practice their tonal words. They submit their audio where we will run it through our machine learning classifier to see if their pronunciation is correct.

How we built it

We used Django as the framework of our web app, Google Cloud SQL for database (not completed), pyAudioAnalysis for building the classifier, and several python and javascript packages to process and display input and output.

Challenges we ran into

Integrating the different components of the project together

Accomplishments that we are proud of

The tone classifier had a 84% accuracy at its height of performance

What we learned

Django, machine learning (supervised classification)

What's next for Tone Sensei

  • Increase accuracy of tone classifier / switch classifier to be based on tensorflow for more customization
  • Solve issues with the linked database
  • Add ranking feature for users to compare their points with their friends on social media
Share this project: