PII detection and redaction in Prompt Engineering / GenAI - Practice Problems & Coding Challenges

Challenge - 5 Problems

🧠 Conceptual

intermediate

What is the primary goal of PII detection in text data?

Imagine you have a document containing names, phone numbers, and addresses. What is the main purpose of PII detection in this context?

A To increase the document's length by adding more details

B To identify and locate personal information so it can be protected or removed

C To translate the document into another language

D To summarize the document's main ideas

💡 Hint

Think about privacy and data protection.

PII detection aims to find personal data like names or phone numbers so they can be secured or removed to protect privacy.

❓ Predict Output

intermediate

What is the output of this PII redaction code snippet?

Given the following Python code that replaces email addresses with '[REDACTED]', what will be printed?

Prompt Engineering / GenAI

import re
text = 'Contact me at [e-mail] or [e-mail].'
redacted = re.sub(r'\b[\w.-]+@[\w.-]+\.\w+\b', '[REDACTED]', text)
print(redacted)

A Contact me at [e-mail] or [REDACTED].

B Contact me at [e-mail] or [e-mail].

C Contact me at [REDACTED] or [e-mail].

D Contact me at [REDACTED] or [REDACTED].

💡 Hint

Look at the regular expression and what it matches.

The regex matches all email addresses and replaces them with '[REDACTED]', so both emails are replaced.

❓ Model Choice

advanced

Which model type is best suited for detecting PII in unstructured text?

You want to build a system that finds names, phone numbers, and addresses in messy text data. Which model type is most appropriate?

A Named Entity Recognition (NER) model trained on labeled PII data

B Image classification model trained on photos

C Clustering model grouping similar documents

D Regression model predicting numerical values

💡 Hint

Think about models that find specific words or phrases in text.

NER models are designed to find and classify entities like names and addresses in text, making them ideal for PII detection.

❓ Metrics

advanced

Which metric best measures the accuracy of a PII detection model?

You have a PII detection model. Which metric tells you how many PII items it correctly finds without too many mistakes?

A F1 score, balancing precision and recall

B Mean squared error

C Accuracy on image pixels

D Silhouette score

💡 Hint

Consider metrics that handle imbalanced classes and both false positives and false negatives.

F1 score balances precision (correctness) and recall (completeness), which is crucial for PII detection.

🔧 Debug

expert

Why does this PII redaction code fail to redact phone numbers?

Review the code below. It redacts emails but fails to redact phone numbers. What is the main reason?

Prompt Engineering / GenAI

import re
text = 'Call me at 123-456-7890 or email me at [e-mail].'
redacted = re.sub(r'\b[\w.-]+@[\w.-]+\.\w+\b', '[REDACTED]', text)
print(redacted)

A The text variable is empty

B The code has a syntax error in the regex pattern

C The regex only matches email addresses, not phone numbers

D The print statement is missing parentheses

💡 Hint

Look at what the regex pattern is designed to find.

The regex targets email formats only, so phone numbers are not matched or replaced.