Enrichment

Turn legal documents into hallucination-free knowledge graphs

Overview

Enrichment

Enrichment is the process of transforming unstructured documents into rich, structured, hierarchical knowledge graphs representing key sections, concepts, entities, and their relationships to one another. It is a critical first step in preparing data for agentic search, eDiscovery, and retrieval-augmented generation (RAG) systems demanding high-quality structured inputs.

Use cases

  • Structured inputs for RAG and search

    Enrichment is a critical first step in preparing data for agentic search, eDiscovery, and RAG systems that demand high-quality structured inputs — turning flat documents into graphs your pipeline can query and traverse.

  • Hierarchy, entities, and relationships

    Capture divisions, sections, clauses, parties, citations, dates, and cross-references — and the relationships between them. Kanon 2 Enricher disambiguates entities, links them together, and deconstructs document structure in one pass.

  • Document intelligence workflows

    Power contract analysis, regulatory mapping, citation networks, litigation workflows, and knowledge bases that need to understand how legal documents are put together — not just what they say in plain text.

  • Hallucination-free by design

    Outputs knowledge graphs, not tokens — so it cannot invent prose, citations, or relationships.

    Many legal AI systems ask a language model to infer structure every time they run. Enrichment moves that work upstream by converting documents into a reusable graph that can be inspected, stored, queried, and connected to the rest of your stack.

    Because Kanon 2 Enricher produces structured graph outputs rather than free-form tokens, it is architecturally incapable of producing the types of hallucinations suffered by general-purpose generative models. It can still misclassify text, but it is fundamentally impossible for Kanon 2 Enricher to generate text outside of what has been provided to it.

  • Purpose-built over generic

    The world’s first hierarchical graphitization model — built for legal documents.

    Unlike universal extraction models such as GLiNER2, Kanon 2 Enricher can not only extract entities referenced within documents but can also disambiguate entities and link them together, as well as fully deconstruct the structural hierarchy of documents.

    It natively outputs to the Isaacus Legal Document Graph Schema (ILDGS) — a first-of-a-kind knowledge graph schema for representing the structure and entities referenced within legal documents.

  • Efficient at scale

    Sub-second latency on documents of any length — small enough to run on a consumer PC.

    Kanon 2 Enricher’s graph-first architecture makes it extremely computationally efficient, outperforming frontier LLMs like Gemini 3.1 Pro and GPT-5.2 on legal enrichment tasks — which suffer from extreme performance degradation over long contexts.

    Kanon 2 Enricher is available for self-hosting via private containers on the AWS Marketplace for enterprises with heightened security or reliability requirements.