Lettria

Overview

Lettria is a document intelligence platform built for regulated industries where accuracy and traceability are non-negotiable. It uses GraphRAG technology to convert unstructured documents into structured knowledge graphs, then answers questions with fully traceable, verifiable references back to source material. The platform targets teams in finance, insurance, healthcare, legal, and life sciences who process complex documents daily and cannot tolerate AI hallucinations. Contracts, claims, medical reports, compliance filings, and other high-stakes documents get parsed, structured, and made queryable without losing the chain of evidence. Lettria''s Perseus model achieves 30% higher accuracy than general-purpose LLMs on graph generation, producing schema-valid knowledge graphs in under 20 milliseconds. Organizations like AP-HP (Paris hospital system) and Leroy Merlin use the platform for patient data structuring and product recommendation improvement respectively.

Features

• GraphRAG engine -- Combines knowledge graph retrieval with generation for traceable, hallucination-free answers
• Document Parsing -- Extracts tables, diagrams, reading order, and multi-column layouts from complex PDFs
• Automatic ontology building -- Generates clean, domain-specific ontologies from your documents without manual mapping
• Text-to-graph conversion -- Transforms raw text into rich knowledge graphs with entities, relations, and constraints
• Entity recognition -- Identifies entities and their relationships across your document corpus
• Multi-language support -- Processes documents across multiple languages for global organizations
• Perseus model -- Proprietary model achieving 30% better accuracy on graph generation with sub-20ms latency

Best For

Financial institutions processing contracts, compliance filings, and risk documents where traceability is mandatory, Healthcare organizations structuring patient data, medical reports, and clinical trial documentation, Legal teams extracting structured information from contracts, regulations, and case filings at scale, Insurance companies automating claims processing with auditable AI that meets regulatory requirements

How It Works

Documents are ingested through OCR and text extraction, handling complex layouts including tables, diagrams, and multi-column pages. Lettria then runs NLP pipelines that perform entity recognition, coreference extraction, syntax analysis, and semantic classification on the extracted text. The structured output feeds into an automatically generated ontology, a domain-specific framework of concepts and relationships. From this ontology, Lettria builds a knowledge graph that maps entities, their attributes, and their relationships across your entire document corpus. GraphRAG combines this graph structure with retrieval-augmented generation. When you query the system, it traverses the knowledge graph to find relevant nodes and relationships, then generates answers grounded in specific source passages. Every answer includes traceable links to the exact document sections it draws from, eliminating hallucinations through structural accountability.

Frequently Asked Questions

What is GraphRAG?

GraphRAG combines knowledge graph traversal with retrieval-augmented generation. Instead of searching flat text chunks, it follows entity relationships in a structured graph to find answers. This produces more accurate, contextual results with full traceability to source documents.

How does Lettria prevent hallucinations?

Every answer is grounded in the knowledge graph, which maps directly to specific passages in your source documents. The system cannot generate claims that are not supported by the graph structure, and every output includes traceable links to its source material.

What industries does Lettria serve?

Lettria focuses on regulated industries: finance, insurance, healthcare, legal, and life sciences. These sectors require auditable, traceable AI outputs that meet compliance standards like GDPR, HIPAA, and SOC 2.

How long does implementation take?

Lettria follows a 12-week deployment process: weeks 1-4 for data connection and initial ontology design, weeks 5-8 for prototype development and testing, and weeks 9-12 for final iteration and deployment preparation.

How does pricing work?

Lettria uses custom enterprise pricing based on document volume, complexity, and support requirements. Contact their sales team for a quote tailored to your organization''s specific document processing needs.