This project tackles bias in text with a two-step approach: generating data and building a smart classifier.
- Data Generation: Using clever prompt engineering, we customized an OpenAI assistant to create a dataset of sentences—neutral or stereotype-laden—by bypassing blocks on explicit stereotype generation.
- Stereotype Classifier: Powered by ALBERT (A Lite BERT), this lightweight yet powerful model detects stereotypes in sentences with speed and precision.
By blending creativity and AI, this tool raises awareness of stereotypes in language, offering a fun yet impactful solution to a serious issue!