Inspiration
Machine learning algorithms discriminate against BIPOC by making wrong predictions in NLP, voice recognition, due to the lack of a diverse training dataset. Many algorithms can be drastically improved if we just had more data—speech, image and text—from BIPOC and minority groups. Biased algorithms can lead to wrong sentiment analysis and inaccurate voice and facial recognition.
What it does
The app allows users to upload audio, text with labels and photos to a shared database. Researchers can download the dataset in order to train their ML algorithms on a more diverse dataset.
How we built it
We built the frontend using React and the backend using Express. The data is stored in an AWS S3 bucket.
Challenges we ran into
As this was our first time using AWS, we had trouble with getting its API working and uploading the files. However, we read many articles and guides online to get them integrated into our app.
Accomplishments that we're proud of
We are proud of building a working prototype in under 24 hours. After facing several challenges like, we were still able to persevere and complete the project we set out to complete.. We take pride in the responsiveness of our application which provides a smooth user interface.
What we learned
We learned research skills, project management and design. We learned about the challenges that BIPOC face and how we can address them through technology.
What's next for TechTogether
We are planning on getting feedback from stakeholders of this project such as researches and BIPOC. With their feedback, we plan on improving the interface of the system and add features. In addition, we plan to add verification of data through randomly selecting users and live training models on the community dataset.
Log in or sign up for Devpost to join the conversation.