Using natural language processing to read online résumés and help employers screen candidates who meet their criteria.
- Built “Résumé Palette” with four teammates: a web system that labels résumé sections automatically and filters résumés by hiring criteria, implemented with Flask and HTML, with data stored in Firebase
- Crawled and de-identified 8,930 résumés and autobiographies from three job sites; five of us labeled 422 of them in Label Studio (9 section types, 3 entity labels) and cross-checked labeling consistency
- Fine-tuned a BERT text-classification model (validation accuracy 0.858) and a question-answering model, filtering résumés by criteria such as certificates and language skills through questions and answers
- Also connected a LINE bot via Dialogflow and Python to search 506 job listings from a job bank; the interface was designed in Figma before implementation
- Won 1st place among the capstone projects of Soochow University’s Department of Data Science (2022)