Moodify

Inspiration

During tough times such as Hackathon's no sleep running or a global pandemic, stress levels tend to rise and affect our daily lives. Moreover, smartwatches are pretty popular nowadays and can monitor and track a comprehensive set of human-physical features and thus, evaluate the stress levels in near real-time. Furthermore, another cause for high-stress levels is news, combined with a crisis… guess what's next?

What it does

We're introducing a new way to experience news reading, taking care of reducing your stress levels. By monitoring the user's data and tracking his human-physical features measured by smartwatches and their preferences, "Moodify" recommends the news suitable to the user's stress levels.

How we built it

We developed a web application and a sophisticated backend using state-of-the-art technology, which involves data analytics, a machine learning algorithm for continuous learning, and profiling user behavior, offering a way to lower its stress levels. Furthermore, NLP techniques, such as sentiment analysis, entities recognition, text similarities to the rescue. We also integrated with given on-site Garmin watch for measuring the person's stress levels.

Challenges we ran into

The big challenge was how to make successful predictions using small, tagged data that we managed to train using the Garmin smartwatch.

Accomplishments that we're proud of

Managed to complete a PoC Involves user interface, backend, and integration with smartwatch API to get reasonable predictions.

What we learned

Working with real-time data, integrating with smartwatch API, and learning algorithms and data visualizations using the d3.js library.

What's next for Moodify

Integration with more smartwatches, going mobile to use the real-time SDK of Garmin, Improving the algorithm's results with more data that we will train continually on more users. Using Reinforcement Learning models for improving personalization.

Built With

Submitted to

HackZurich 2021

Created by

I worked on the back-end, and I implemented a sentiment analysis system and calculated the similarity between the articles with python and IBM tools. In addition, I was working on a machine learning model that recognizes the user's mood from his measurement from the smartwatch and his feedback after using the app.

Mira Finkelstein
UI/UX and data prep

Bar Haim
- backend, implemented using python Flask.
- Established the connection to the Garmin watch using the fitrockr api
- Explored options with working serverless on Azure and other alternatives.
- Managed an Azure subscription to host our app on Azure app
- Created a GitHub ripo and connected it to the app service to try and have the fastest delivery possible during continuous development at the Hackzurich.

Rom Cohen
Eliyahu Strugo