Logo and App Design for Code for San Francisco
Problem: Computers aren't great at telling the difference between a number that refers to a bus route and a number that refers to a time, on their own. They lack nuance, and AI needs good datasets to train on.
Labeling the contents of thousands of tweets by hand is tedious. How might we make it faster and easier, while reducing cognitive load?
Solution:A labeling method for ML NLP training sets that improves on old methods by segmenting and batching tweet labeling tasks, so that labelers can spend less time thinking and more time labeling.
Role:Volunteer Designer
Impact:Designed new and more efficient tweet labeling method using Figma and created a clickable prototype demonstration that was able to reframe tweet-labeling tasks for the team.