Classification

Label legal documents zero-shot, with no training data

Overview

Legal labels without training data.

Classification determines whether a natural-language statement about a document is supported by that document. Use it to label clauses, cases, evidence, policies, and other legal materials without collecting examples, finetuning a model, or building a classifier from scratch.

Why it matters

Turn legal judgment into confidence scores

Kanon Universal Classifier gives legal engineers and product builders a fast, accurate way to test legal statements against documents at scale.

  • Label anything, zero-shot

    Classify legal documents, clauses, passages, or evidence against natural-language statements with no training data, finetuning, or bespoke taxonomy required.

  • Fast enough for large corpora

    Kanon Universal Classifier can evaluate a statement against thousands of documents in mere seconds, making classification practical for retrieval, review, and extraction at scale.

  • Tiny but mighty

    Despite their compactness, Kanon and Kanon Mini punch far above their weight, achieving 6% and 12% better performance than their closest general-purpose counterparts.

How teams use it

From document piles to product decisions

Build review tools, triage systems, risk classifiers, retrieval filters, and legal workflows that can classify documents as quickly as your product can send them.

  • Zero-shot legal classification

    Turn legal judgment calls into fast, reusable classification checks without collecting training data first.

    Legal teams classify documents constantly: whether a clause creates a termination right, whether a document contains privileged material, whether a judgment applies a doctrine, or whether evidence supports a factual proposition. Classification turns those decisions into scalable product primitives.

    With Kanon Universal Classifier, you can write the label as a natural-language statement and evaluate it against documents directly. Ask whether “this clause entitles one to terminate an agreement in the event of circumstances beyond their reasonable control” is supported by a document, and receive a confidence score in return.

    For legal engineers and vibe coders, the workflow is simple: describe what you want to detect, send the documents, and use the scores to label, filter, route, rank, or trigger the next step in your application.

    • Zero-shot
    • Legal QA
    • Document review
  • Kanon Universal Classifier

    The world’s most accurate and efficient universal legal classifiers of their size.

    Kanon Universal Classifier and Kanon Universal Classifier Mini can take a statement like “this clause entitles one to terminate an agreement in the event of circumstances beyond their reasonable control” and evaluate it against thousands of documents in mere seconds, producing startlingly accurate confidence scores — no finetuning necessary.

    Despite their compactness, Kanon and Kanon Mini punch far above their weight, achieving 6% and 12% better performance, respectively, than their closest general-purpose counterparts.

    Use them anywhere you need fast legal judgment at scale: issue spotting, clause detection, privilege review, risk tagging, matter triage, regulatory mapping, evidence analysis, or filtering retrieval results before a downstream model sees them.

    • Kanon
    • Universal classification
  • Classification as a product primitive

    Use confidence scores to build legal workflows that route, rank, filter, and act on documents automatically.

    Classification is useful when the product needs a decision, not a paragraph. Is this a confidentiality clause? Does this contract contain a force majeure provision? Is this authority about proportionality? Does this evidence support the allegation?

    Because Kanon Universal Classifier works zero-shot, teams can iterate on labels as quickly as they iterate on product ideas. You do not need to collect examples, train a classifier, or maintain a separate model for every new workflow.

    Combine classification with embedding, reranking, extraction, or enrichment to build richer legal systems: retrieve a candidate set, classify it against the legal issue, extract the answer, and structure the result for review or automation.

    • Risk tagging
    • Matter triage
    • Workflow automation