Rasa Whiteboard: NER for PII Detection Challenges

Detecting personally identifiable information from text is typically a NER problem. It's an important problem, especially when you're thinking about labeling pipelines, but it's also very hard to get right. The goal of this video is to explain why it's such a hard problem and we'll poke at some systems to see how they might fail.

If you'd like to learn more about Presidio, you can find the project on GitHub here:
https://github.com/microsoft/presidio

You can find the live demo used in this video here:
https://presidio-demo.azurewebsites.net/

You can find more information about the equity evaluation corpus here:
https://saifmohammad.com/WebPages/Biases-SA.html

You can play with the `ner-english` Flair model on huggingface here:
https://huggingface.co/flair/ner-english