Text Classifier

Incoming text. A useful place to start.

Messages arrive with different needs. Put the most relevant labels in front of the people and systems deciding what happens next—without training a custom classifier.

“I was charged twice.”

Billing

“I can't sign in.”

Access

“My export won't finish.”

Technical support

Illustrative top matches shown. Full results rank every supplied label.

From message to likely destinations

Same labels. A different order.

Choose a sample message and see the destinations reorder. All supplied labels remain in the result.

Incoming message

I was charged twice for my subscription. Can you help me get the second payment back?

Sample A · Payment question

I changed my phone and can't sign in. How do I get back into my account?

Sample B · Access question

My report export stops halfway through every time. Can you help me get the file?

Sample C · Product issue
Your labels, ranked by similarity
  1. 01
    Billing & paymentsCharges, invoices, refunds
  2. 02
    Account accessSign-in and account recovery
  3. 03
    Technical supportProblems using the product

Worked example · illustrative ranking, not a measured result · no live call. Classifier returns similarities; it does not execute routing.

Illustrative sorting machine separating paper slips into different trays
Bring your vocabulary to the work—not categories invented for you.

Less undifferentiated text

Make a growing inbox easier to review.

  • Support triagePut likely queues in front of an operator.
  • Content organizationCompare articles and feedback against your topics.
  • Repeated workUse one label set across multiple documents, with results matched to each input.

A useful signal. Not a verdict.

Classifier compares the meaning of your text and labels. Similarity is not a confidence percentage or a guarantee of correctness. Your application decides whether to route, review or ask for more context.

Need to ask explicit questions?

Choose Classifier for label similarity. Choose Decisions API for named yes/no, choice and scale judgments with instructions and criteria.

Explore Decisions API

Bring messages your team already understands.

Start with a focused label set and known examples. Compare the returned rankings before automating a destination.

A lightweight processing step.

Classification uses embedding-token billing and document quotas. Every document-label pair is returned, so keep the category set purposeful.

Text tool pricing and limits