Integrating Website Classification API into Crime & Justice Workflows

You need to automatically classify domains that touch crime and justice topics so you can block them for brand safety, route them to a specialist review queue, or enrich signups with context before they hit your CRM. By the end of this guide, you’ll send a URL to the Klazify categorize endpoint, read the category fields you need, and wire a yes/no decision into your pipeline—complete with caching and fallback handling for unknown domains.

Why Klazify fits crime and justice workflows

Teams working with crime and justice signals care about precision, speed, and structured categories they can act on. Klazify’s categorize endpoint is designed for this:

• Accurate website categorization using AI: It analyzes full-page content, enabling reliable classification even when crime or justice topics are present in news sections, press releases, or subpages rather than just meta tags.
• Global coverage: Crime and justice content is multilingual and distributed across international government portals, NGOs, and media; Klazify supports classification across languages so you can apply a single policy globally.
• Real-time classification: Crime and justice pages change quickly as news breaks and policies evolve. Klazify emphasizes fresh analysis instead of outdated snapshots.
• Industry-level categories via IAB taxonomy: Because the API maps to IAB taxonomy, your ad tech and brand safety controls can stay aligned with an industry standard, and you can keep a consistent policy across channels.
• Simple API integration: A single POST request to the categorize endpoint returns categories, logo, similar domains, and company context you can use for enrichment or filtering.
• Compliance and filtering support: Use confidence scores and hierarchical category names to implement precise allow/block rules for law enforcement topics, court resources, legal news, or other justice-related segments.

Below, you’ll see a complete flow to detect and route crime and justice domains in production using Klazify’s Documentation and the categorize endpoint.

Scenario: brand safety blocklist, signup enrichment, and traffic auditing

Here are three concrete ways teams wire Klazify into crime and justice workflows:

• Brand safety blocklist: Before your DSP or content network serves an ad, look up the landing page URL. If categories match your internal “crime & justice” mapping (e.g., law enforcement agencies, courts, prison services, legal policy), exclude placements or reroute to alternative creatives.
• Signup enrichment: When a user signs up with a corporate email or website, call categorize on their domain. If the categories map to your justice-related verticals, enrich their CRM record with tags for sales routing (e.g., public sector, legal services) without waiting for manual research.
• Ad auditing: Periodically crawl where your ads appeared, classify each publisher or page URL, and validate none fall into restricted crime/justice segments per campaign rules.

Send a POST /api/categorize request

The categorize endpoint accepts a URL and returns categories, company data, logo, and related signals for decisioning. Authentication uses a Bearer token in the Authorization header.

curl -X POST "https://www.klazify.com/api/categorize" \
-H "Authorization: Bearer YOUR_API_KEY" \
-H "Content-Type: application/json" \
--data '{"url":"https://www.justice.gov"}'

Notes:

• Replace YOUR_API_KEY with your token after you Register.
• Use a fully-qualified URL including protocol. You can send homepages or deep links for page-level classification.

The categorization response (official example and field guide)

Below is the official example response format you’ll receive from the categorize endpoint. Use this structure to parse categories, confidence scores, IAB mappings, logo URL, and company objects.

{
"domain": {
"categories": [
{
"confidence": 0.92,
"name": "/Computers & Electronics/Consumer Electronics",
"IAB-632-596": "Consumer Electronics/Technology & Computing/Consumer Electronics"
},
{
"confidence": 0.89,
"name": "/Internet & Telecom/Mobile & Wireless/Mobile Phones"
}
],
"social_media": null,
"logo_url": "https://klazify.s3.amazonaws.com/2110787991611585019600ed5fb1d1300.04730104.png"
},
"success": true,
"objects": {
"company": {
"url": "https://www.apple.com/",
"name": "Apple",
"city": "Cupertino",
"stateCode": "CA",
"countryCode": "US",
"employeesRange": "100K+",
"revenue": 274515000000,
"raised": null,
"tags": [
"E-commerce",
"Consumer Electronics",
"Mobile",
"B2C"
],
"tech": [
"omniture_adobe_analytics",
"atlassian_confluence",
"successfactors",
"apache_apex",
"talend",
"oracle_peoplesoft",
"salesforce",
"stripe",
"dell_boomi_atomsphere",
"gigya",
"sage_50cloud",
"quickbooks",
"webmethods",
"apache_tomcat",
"alteryx",
"tibco_rendezvous",
"atlassian_jira",
"..."
]
}
},
"domain_registration_data": {
"domain_age_date": "1987-02-19",
"domain_age_days_ago": "13026",
"domain_expiration_date": "2030-02-20",
"domain_expiration_days_left": "123"
},
"similar_domains": [
"bestbuy.com",
"icloud.com",
"microsoft.com",
"macrumors.com",
"google.com",
"samsung.com",
"twitter.com",
"hp.com",
"bhphotovideo.com",
"dell.com"
]
}

How to use these fields for crime and justice workflows:

• domain.categories: The primary data for allow/block or routing. Use name for hierarchical matching and confidence to gate actions (for example, block only if confidence ≥ a threshold). If present, IAB-632-596 provides a direct IAB mapping string you can use in ad tech pipelines.
• domain.logo_url: Good for enriching CRM or internal admin tools so analysts can quickly recognize a site.
• objects.company: Use url and name for display, tags for enrichment, and location fields for routing to regional teams. tech can hint at the sophistication or integrations used by the domain.
• domain_registration_data: Useful for risk signals, e.g., very new domains may deserve manual review before allowing ads or traffic.
• similar_domains: Seed expansion. If a government justice portal is allowed, you may want to fast-track classification of related official domains; if something is blocked, you can proactively evaluate similar domains.

Mapping categories to crime and justice policies

Your internal policy likely doesn’t match category names one-to-one. The typical approach is to define a mapping layer from category strings to an internal decision. Since category names are hierarchical strings, substring and prefix matches are useful. Keep a normalized list of terms you consider “crime & justice” (for example: “Law”, “Legal”, “Government”, “Public Safety”, “Courts”, “Corrections”, “Police”).

Because category values are returned as readable names and, where available, IAB mapping strings, you can:

• Use name to perform prefix or keyword matching across hierarchies.
• Use the IAB mapping string when you prefer standard taxonomy alignment in ad tech workflows.
• Combine both with confidence to produce a “yes/no/needs review” decision.

Python example: classify and decide yes/no for crime & justice

The example below sends a URL to categorize and maps categories to a yes/no decision. Adjust the keyword list and confidence threshold to your policy.

import os
import requests

API_KEY = os.getenv("KLAZIFY_API_KEY", "YOUR_API_KEY")
ENDPOINT = "https://www.klazify.com/api/categorize"

# Keywords you consider to indicate crime & justice topics.
# Tune this list to your policy and taxonomy.
CRIME_JUSTICE_KEYWORDS = [
"Law", "Legal", "Government", "Public Safety", "Police",
"Court", "Courts", "Corrections", "Judicial", "Justice", "Prosecution"
]

CONFIDENCE_THRESHOLD = 0.75 # Adjust to your risk tolerance

def is_crime_justice(categories):
if not categories:
return False
for c in categories:
name = c.get("name", "") or ""
iab = c.get("IAB-632-596", "") or ""
conf = c.get("confidence", 0.0) or 0.0
# Require confidence above threshold for a match
if conf >= CONFIDENCE_THRESHOLD:
haystacks = [name, iab]
for h in haystacks:
for kw in CRIME_JUSTICE_KEYWORDS:
if kw.lower() in h.lower():
return True
return False

def categorize(url):
headers = {
"Authorization": f"Bearer {API_KEY}",
"Content-Type": "application/json"
}
resp = requests.post(ENDPOINT, json={"url": url}, headers=headers, timeout=20)
resp.raise_for_status()
return resp.json()

if __name__ == "__main__":
test_url = "https://www.justice.gov"
data = categorize(test_url)

domain = data.get("domain", {})
categories = domain.get("categories", [])
decision = "ALLOW"
if is_crime_justice(categories):
decision = "BLOCK"

print("URL:", test_url)
print("Categories:", categories)
print("Decision:", decision)
print("Logo:", domain.get("logo_url"))
print("Company:", (data.get("objects", {}) or {}).get("company"))

Use this as a starting point. If you rely on IAB mappings, populate CRIME_JUSTICE_KEYWORDS with strings that correspond to the IAB paths you consider related to justice topics, and rely primarily on IAB-632-596 when present.

How to act on the returned fields

• Blocking (brand safety): Evaluate domain.categories[].name and domain.categories[].confidence. If any category matches your mapping at or above a threshold, block the placement or page. Log the matched category and confidence for audits.
• Signup enrichment (CRM): Use objects.company fields to fill in name, location, and tags. Use domain.logo_url for UI displays. Store categories so SDRs see the domain’s focus at a glance.
• Auditing: Store the full JSON for evidence. The domain_registration_data block is helpful to surface new or expiring domains during weekly reviews.

Field-to-use-case quick reference

Operational details that save time in production

Caching per domain

Cache results keyed by the normalized registrable domain (e.g., example.gov). For sites with highly variable subpages, also cache per URL for page-level classification. Suggested approach:

• Cache TTL: Set a TTL aligned with how fast your vertical changes. Many teams start with hours to days. Shorten TTL for news-heavy publishers.
• Version the cache entry by URL vs. domain so you can choose the best level per integration (brand safety often wants page-level granularity).

Handling unknown or new domains

• First-seen workflow: If a domain has no cached entry, call categorize and hold the decision until it returns or fails over to a conservative default (e.g., temporary block or review).
• Fallback: If the response has empty categories or low confidence, treat it as “needs review” or allow per your risk tolerance.
• Warm-up queue: Periodically scan your logs for high-traffic new domains and pre-classify them to avoid cold-start latency.

Batching and throughput

• Batch in your client: Group URLs and process them concurrently with a pool (size tuned to stay within your account’s allowed throughput).
• Backoff: On HTTP 429 or similar, apply exponential backoff and retry with jitter. Do not guess rate limits; observe responses and consult the Documentation for account-specific guidance.
• Idempotency: Since classification is deterministic for the same URL at a point in time, retries can safely re-run without side effects on your store.

Mapping to your taxonomy

• String maps: Maintain a list of keywords and prefixes that represent your “crime & justice” concept. Include common variants and synonyms your team uses.
• IAB alignment: Where the response includes IAB mapping strings, prefer matching on those for ad tech pipelines. Keep a compact lookup table from IAB strings to your internal labels.
• Confidence-aware rules: Combine category hits with confidence thresholds. Example: auto-block when confidence ≥ 0.85; send to review for 0.6–0.85; allow below 0.6.

Storing and auditing results

• Persist the raw JSON for at least the lifetime of your decision (e.g., the ad flight or the CRM lead lifecycle) so you can reproduce and defend actions later.
• Log matched category names, IAB mapping strings, and confidence scores alongside your yes/no decision.
• Track drift: Re-run classification on a schedule for high-traffic domains and compare results. If categories change meaningfully, trigger a re-evaluation of policy decisions.

End-to-end decision flow

1. Receive a URL (e.g., a publisher page for ad placement, or a domain from a signup form).
2. Normalize and check the cache. If fresh, use it; otherwise, call POST /api/categorize.
3. Parse domain.categories. If present, evaluate each name and (if available) the IAB mapping string against your internal “crime & justice” mapping with a confidence threshold.
4. Produce a decision: ALLOW, BLOCK, or REVIEW.
5. Enrich records: Store logo_url and objects.company.* where applicable.
6. Record audit data: Save the JSON, matched rule, and the final decision.

Practical implementation notes

• Character encoding: Treat category names as UTF-8 strings; normalize casing for keyword comparisons.
• Timeouts: Set HTTP client timeouts. The Python example uses 20 seconds; tune per your SLA.
• Resilience: Implement retry logic for transient failures and network errors. Avoid unbounded retries.
• Security: Store your API key securely (environment variables or your secret manager). Never hardcode keys in client-side code.
• Page vs. domain scope: Use page-level classification for mixed-content sites (e.g., media) and domain-level for uniform sites (e.g., government portals).

Where MCP fits

If you manage centrally-governed policies, review the MCP page to understand how managed configuration and policy frameworks align with your deployment. This helps keep your allow/block strategy synchronized across teams, services, and environments.

Common pitfalls and how to avoid them

• Only checking the first category: Always scan all items in domain.categories. A lower-ranked category can still be decisive.
• Ignoring confidence: Combine category matches with a confidence threshold to reduce false positives and avoid over-blocking.
• Assuming social links: domain.social_media can be null. Don’t rely on it for routing decisions.
• Static mappings: Refresh your mapping dictionary as your policy evolves. Keep it versioned and testable, just like code.
• Skipping rechecks: Content changes. Schedule reclassification for high-impact domains and near-expiry registrations.

FAQ

Can I send page URLs, or should I only send the root domain?
You can send either. For mixed-content publishers, send the specific page URL to get page-level categorization. For uniform sites (e.g., official justice portals), domain-level may be sufficient.

What if the categories array is empty or confidence is low?
Treat it as “needs review” or use your conservative default (e.g., temporary block). You can also retry later or after content changes.

How should I cache results?
Cache per URL for page-level decisions and per registrable domain for domain-level decisions. Set a TTL appropriate to your use case and refresh proactively for high-traffic or high-risk sites.

How do I align with my ad tech taxonomy?
Use the IAB mapping string (when present) from domain.categories to match your ad tech policies. Maintain a lookup from IAB strings to your internal labels and evaluate with a confidence threshold.

How do I avoid hitting throughput limits?
Process URLs in bounded parallel batches, implement exponential backoff on 429s, and pre-classify frequent or high-impact domains. Consult the Documentation for details.

Get started by creating a free account and wiring the categorize endpoint into your pipeline. Build your crime and justice decision layer, cache smartly, and ship with confidence: Register. For full endpoint specs and field definitions, visit the Documentation.