Inspiration
AI is a territory both of us wanted to delve into. We are also passionate in communities such as suicide prevention. Having had personal experience with the darker topics of depression, self-harm, and suicidal thoughts and tendencies, it has become a passion to help others when possible, and to ensure that people know that they are cared for, and where they can seek help. This application was the first experiment into aiding the community with awareness and encouragement.
What it does
Awake, Alive is a self-help AI to handle basic conversation between the user and machine. It uses a contextual dataset to render responses for the user. Awake, Alive deals with difficult topics, such as depression and abuse. This application was designed to give the user someone, or in this case something,to vent to, and to possibly receive small tips of advice as well. Awake, Alive focuses on the safety of the user, and encourages them to seek help if possible.
How we built it
We created a simple neural network, trained it, then created a Flask front-end. We did more training and added data to our contextual AI dataset.
Challenges we ran into
With neither of us having experience in Tensorflow, the initial understanding took the most amount of time. We also found issues in modules not being installed, and in javascript taking time to learn communication between the json and card list with bootstrap.
Accomplishments that we're proud of
We're proud of having the front up and communicating. This was a challenge, and neither of us had experience with Tensorflow. Being able to see that first ai response was the best accomplishment of the hackathon, let alone getting the application fully up and communicating.
What we learned
We were both new to AI, and had never used Tensorflow. This was a start-to-finish learning curve on how to teach and use AI.
What's next for Awake, Alive
We would like to adopt this software into a personal project we have been working on. AI will be a large factor in this project, and we are thrilled to have more experience. We also look forward to retrieving larger datasets for more variety of topics and responses, as well as learning to web-scrape for these sets.
Log in or sign up for Devpost to join the conversation.