Inspiration
Our team had an interest in building a webapp that was powered by AI. We decided to look for ways we could use data to build a model to counteract racial discrimination.
What it does
Given a text query, our model predicts a score based on how likely it is to be racially biased, hate speech, or otherwise discriminatory in any way.
How I built it
Using a formatted dataset of over 30,000 entries, we trained a machine learning model using TF-IDF for the document term matrix and a Bagging Classifier as the ML model. We also integrated Google Cloud's AutoML to add portability to the application by being able to predict on the cloud.
Challenges I ran into
Connecting our Python-built machine learning model to a webapp proved to be a difficult process. We ended up using two separate web servers for backend and frontend communication.
Accomplishments that I'm proud of
Being able to achieve somewhat practical levels of accuracy with our model given the limited experience is probably our biggest accomplishment.
What I learned
Different ways to format data for ML, models for training and methods of evaluation. Additionally, learned about Google Cloud's suite of products as well as understanding of WebAssembly and its implications.
What's next for Racial Bias AI
Improving the model and making a proper deployment, potentially featuring a public API, which can be used by anyone to filter out racial discrimination.
Log in or sign up for Devpost to join the conversation.