Posi-Flower

• 

  This is the message received from the online AI chatbot, which is trained by parlai model.

Inspiration

In a world where covid19 runs rampant, the word "positive" lost its initial glow and instead became a word that alarms. This project is bringing back the true meaning of positive.

Deterioration of mental health is an important global issue that affects everyone, yet it isn't served as much attention as it should. We wanted to give back to the world that raised us by taking away some stress from everyone and helping improve the mental health of those in need.

What it does

There are two ai models in this app for two main features. The first is a conversational ai agent that maintains a conversation with the user, this allows the user to rant about anything they want to. The second is an NLP model that applies sentiment analysis on the text messages of the user. If the NLP model outputs positive numbers meaning they detected positivity in their message, then an animated sunflower will grow a level. This is done in hope that the user gets motivated in writing positive comments.

How we built it

We built this posi-flower app in Java programming language using Android Studio. The animations and images in the GUI interface of the app were drawn by hand frame-by-frame using Flipaclip. This drawing style is chosen to make the interface more adorable and happy.

Challenges we ran into

Facebook Research proposed a very powerful chatbox machine. However, we experienced difficulty fitting this model properly into the android studio interface. We initially considered uploading the ML model on google cloud so that it can be served to the user, but due to time constraints, we could not get the Java application to work with the ML model.

We chose to implement a simple reflection to do the user-machine interaction, instead of making the model for real-time interaction.

Accomplishments that we're proud of

We are proud of the java application that we ended up with and the cute animations and images in the user interface. In our design of this project, it was desirable to add a human element in this app as it was already lacking so much.

What we learned

On Feb 5th, four strangers came together to form an ad-hoc team. We learned so much in a short time, including aspects of software engineering, UI design, Machine Learning, but especially how to work in a team with people you never met before. Everyone agrees that it was a rewarding experience! This event also offered great opportunities to meet inspiring sponsors. The challenge our team chose to tackle was one that made us think a lot. We were really grateful for what the world has offered us so far, and being a member of this full-of-crises world, we decided to do something in return and do our best to help. Although some of our team members have the experience of constructing an android app, the rest of us gained valuable software engineering experience. We also learned that sometimes it is necessary to give up the more elegant solution path and instead take the shorter paths with patchy solutions.

What's next for Posi-Flower

few features we want to add in the future if we have the time:

1. introduce different conversational ais for the user to choose from. (We want it to make the experience of talking to AIs as transparent as it can be, instead of fooling the user to think that they are talking to a real person. Having different options helps the user recognise that)
2. use a better production system (maybe Kubernetes and use the google cloud)
3. plug in a lifeline that allows the user to call a helpline when the sentiment analysis detects severe detrimental language. ( or if the user shows signs of suicidal thinking)

Built With

• java
• parlai
• python