Inspiration
In the post COVID era humanity has realised the consequences of a health crisis. From economical to more importantly physical and mental. This newfound and re-instilled awareness and importance to health has rightfully motivated us as a species to be as healthy as possible. The question is where does good health begin? rather than treating diseases at hospitals as they develop, a healthy lifestyle and nutritious diet prevents us from being in immunocompromised positions. Furthermore, misinformation about nutrition and diet is constantly circulating the web leaving users wrongly educated about their nutrition and diet needs.
The human requirements of nutritious diets and healthy lifestyles for optimal living inspired us to make HealthSimple, to keep your nutrition and diet needs simple.
What it does
HealthSimple empowers users towards healthier lifestyles and nutritious diets through the power of knowledge and awareness. Here's the users execution walk through:
The user wants to eat a tasty meal while being informed about their diet choices
The user enters the individual ingredient components of their meal into our app (ex. Chicken, Broccoli, Rice)
HealthSimple provides the user with a meal summary and grade indicating the nutritional quality of this meal
HealthSimple also provides nutritional benefits information about each ingredient in a simplified and summarised paragraph from reliable and trusted sources
How we built it
HealthSimple is a React(frontend) + Flask(backend server) application. The summarization and meal quality text generation capabilities are powered by NLP with Coheres API.
Challenges we ran into
We were not able to find datasets to train our custom Cohere API Generate model, so the whole team banded together to manually create a custom dataset to train the article summarization model. This was our first time web scraping and we had a lot to learn, which posed its own challenges. We decided to web scrape nutrition articles from webMD, but our attempts were blocked. We considered using a proxy to overcome this, however we later learned that we could use a certain web scraping API with an inbuilt proxy. Another challenge we ran into was finding a nutrition API to suit our needs. We spent a considerable amount of time looking through documentation for various nutrition API’s to decide what works best for us. We ended up using Edamam which did not provide us with all the necessary micronutrient data we were looking for. We had to accordingly update our health grading algorithm in order to use the data we have.
Accomplishments that we're proud of
We are proud that we were able to build a working web app. We are also proud of the fact that we were able to find multiple creative ways to implement NLP models to elevate user experience.
What we learned
What's next for HealthSimple
We would like to deploy this app for users to be able to start using HealthSimple immediately. We would like to be able to use more scientifically accurate data to calculate the nutrition grade by using the peer - reviewed nutrition API : Nutri-score.
Log in or sign up for Devpost to join the conversation.