Pathway

Inspiration

In 2018, around 39% of the population stated that they didn’t seek mental health because it was too costly. People are unable to find proper therapy because of lack of proper information and oftentimes simple notifications can provide the motivation for someone who is down. Cost acts as a barrier between those that can obtain medical assistance and those that can’t. We realized all of this when many people at and outside of our schools sought therapy for many different reasons, but they had regrets and hesitance while doing so because of the cost. Especially in the midst of a pandemic, the financial conditions of many families became fragile. We wanted to make it a leveled field for everyone because the right to life should not be decided by wealth. People have also lost time and money by trying to solve their needs through therapy.

What it does / How We Built It

nce you enter our app, you are prompted with a welcome screen where you can select whether you wish to create a new account or log into an existing one. If you want to create a new account, you are prompted with an email, name and password input for your account. This data is then stored into Google Cloud’s Firebase where all the users’ data is stored. From there, you are logged into the app where you can do one of our many features. Once you log in, you can view the dashboard page, where you can see your customized “Daily Quote Board” where you can add quotes for other people to read or create them for yourself for later use. The purpose is to make the user and other users feel better about themselves. Each quote that is put into the app is immediately stored in a secure database provided by Google, Firebase Cloud Firestore. It allows for easy retrieval and modification of information. Personal quotes can also be generated by selecting a category (Love, Sports, Life, etc.), and a quote from a famous person will appear on the screen. If you go to the “Therapy Finder”, you can see different therapists within a 25 mile radius that you can contact to get medical assistance for your needs. Our app provides a platform where you can call, message, email, and/or book an appointment with them. This allows for users to save time on their own by preventing them from scouring the internet and finding one result. Now, our primary feature of the app is the Therapy Recommender. The functionality of the feature is very similar to how Netflix recommends certain movies and shows for its users to watch. You would type your concern or how you’re feeling at the moment in the search box, and with machine learning using Keras; Tensorflow; Flask; Python, we can provide recommendations on what would be the most beneficial option for the user. Once the user submits the information, the app will provide different links for the user to choose from which gives them resources on what to do. They can rate the recommendation, submit feedback, and open the source from the app if they would like. By rating the recommendation, it will allow for the backend to learn what suggestions are more helpful for users, allowing it to give more customized feedback.

Challenges we ran into / What we learned

Integration of our algorithm also proved challenging and difficult to integrate and learn, in which the use of sentiment analysis, a word to vector model, and cosine similarity to create a precise way to diagnose the patient. This was extremely difficult and had to be thoroughly researched and well thought out before adding it to the code. The content-based recommendation system also contained many difficulties. In order to store all the ratings, we had to use a cloud-based database: Google’s Firebase. Learning the intricacies of interacting with this service was extremely time consuming. Understanding how to model the data in matching format for the algorithm and the creation of an API through Flask took a great deal of time. There is a common issue in recommendation systems called the cold start problem. The cold start problem is a problem which arises because of lack of meaningful data. How can we make recommendations without the initial presence of data? In order to solve this, we decided on a content-based system with manually imputed initial data. To sum it up, the creation of the app and especially some of its technical aspects were difficult, but we were able to overcome and learn from them.

What's next for Pathway

If given time to create a 2.0 version of the app, we plan to add more refinements in our algorithms to further improve suggestions in our app, such as adding more categories which would use a greater dataset for more specific suggestions to increase accuracy and precision. We then would build a platform to connect therapists virtually to provide human assessment to confirm potential diagnoses they may see or schedule a simple chat to expand the app’s impact and promote mental and physical health.

Built With

• flask
• javascript
• keras
• python
• react-native
• tensorflow