Privacy Protection Toolkit for Video and Transcript Data: Over-redaction Tool

Code/Algorithm

Open-source web application for CSV file de-identification and redaction review, combining AI-powered PII removal with manual review capabilities.

Visit Resource
This link will take you to an external website.
Purpose/Abstract

The RPPL Privacy Toolkit is an integrated web application for CSV file de-identification and redaction review.

The toolkit combines AI-powered personally identifying information removal with manual review capabilities, supporting privacy-preserving workflows for educational data.

The application allows users to upload and process multiple CSV files, select which columns to de-identify, automatically generate ID columns when needed, and process multiple text columns in a single run.

It includes a visual diff interface that enables users to review redactions and undo specific changes when the de-identification process has over-redacted information.

The toolkit also supports ground-truth evaluation, allowing users to compare results against human-validated redactions and calculate accuracy, precision, recall, and Kappa scores.

The application can be run with OpenAI models or configured to use local language models through tools such as Ollama or LM Studio, supporting different privacy and deployment needs.

The resource is designed to increase efficiency, support secure data sharing, and help researchers and developers process sensitive educational data while preserving privacy.

Citation
RPPL Privacy Toolkit.

Areas researched: Data Use, AI

Previous
Previous

Informing Change: A Final Research Report on the Findings from the Longitudinal R&D Partnership for Math Equity Project

Next
Next

RPPL Visualizer and Converter