Pin Violation Analysis - Detect and Categorize Policy-Violating Pins

Question

## Round 1 - Coding

## Problem

You are given a list of pins, each with metadata (title, description, tags, image_labels). You also have a policy ruleset specifying forbidden keywords and label combinations. Identify all violating pins and classify the violation type.

```python
def analyze_violations(
    pins: list[dict],
    policy: dict
) -> list[dict]:
    # pins[i] = {"id": int, "title": str, "tags": list[str], "labels": list[str]}
    # policy = {
    #   "forbidden_keywords": list[str],
    #   "forbidden_label_combos": list[list[str]]  # flag if ALL labels in any combo present
    # }
    # Returns: [{"pin_id": int, "violation_type": str, "details": str}]
    ...
```

```
Example:
pins = [
    {"id": 1, "title": "best weapons guide", "tags": [], "labels": ["firearm", "violence"]},
    {"id": 2, "title": "healthy recipes",    "tags": [], "labels": ["food"]},
]
policy = {
    "forbidden_keywords": ["weapons"],
    "forbidden_label_combos": [["firearm", "violence"]]
}
analyze_violations(pins, policy) -> [
    {"pin_id": 1, "violation_type": "keyword",   "details": "weapons"},
    {"pin_id": 1, "violation_type": "label_combo", "details": "firearm+violence"}
]
```

## Follow-ups

1. How would you handle case-insensitive and partial keyword matching efficiently?
2. A pin can have multiple violations - how does your output model handle that?
3. How would you build an audit trail so removed pins can be reviewed?
4. What false-positive risk do keyword filters carry and how do you mitigate it?

Pin Violation Analysis - Detect and Categorize Policy-Violating Pins

Question Details

Full Details

Topics

More from Pinterest