Introduction
Learn how Perseus, Lettria's Text-to-Graph engine, transforms unstructured text into structured knowledge graphs for your AI systems.
From Unstructured Text to Actionable Knowledge
Perseus is Lettria’s Text-to-Graph engine. It’s designed to solve a fundamental problem modern AI faces: most organizational knowledge is locked away in unstructured text, while AI systems thrive on explicit, connected data.
Perseus bridges this gap. It transforms your documents, articles, and reports into a structured knowledge graph that your AI systems can reliably understand, query, and reason over.
The Challenge: The Limits of Raw Text
Unstructured text is chaotic. Information is hidden, relationships are implicit, and getting reliable answers is difficult. In contrast, AI systems need clear, structured, and connected data to perform effectively.
Before Perseus: Unstructured Text
- Entities are implicit and ambiguous
- Relationships are scattered across documents
- Queries rely on brittle keyword search
- Reasoning across documents is unreliable
After Perseus: Structured Knowledge
- Entities are clearly identified and typed
- Relationships are explicit and queryable
- A consistent schema enforces structure
- Data can be traversed, filtered, and analyzed
Perseus converts this hidden information into connected knowledge, making it usable far beyond simple text retrieval.
The Solution: A Text-to-Graph Engine
At a high level, Perseus provides a clear, repeatable process for knowledge extraction.
Analyze Raw Text
Perseus ingests your source documents and performs a deep linguistic analysis to understand their content.
Extract and Map to a Schema
It then identifies and extracts key entities and the relationships between them, mapping them to a structured schema you provide—an ontology.
Output a Knowledge Graph
The final output is a clean, structured knowledge graph, ready to be loaded into a database or used directly in your AI application.
The Core Feature: Ontology-Driven by Design
A key strength of Perseus is that it is ontology-driven.
Instead of producing a generic graph, Perseus uses an ontology that you define. This allows you to control exactly what gets extracted and how it is structured, ensuring the output is tailored to your specific domain. Your ontology becomes the contract between raw text and structured knowledge, reducing noise and ambiguity.
This makes Perseus especially powerful for enterprise knowledge bases, domain-specific AI assistants, and any application requiring high-precision, structured data.
How Perseus Empowers Your AI
By integrating Perseus into your AI stack, you give your applications a reliable knowledge layer to build upon, moving beyond simple keyword search or embeddings. The SDK provides the tools to programmatically control this entire process.
Perseus is the right choice if you need:
- Explainable, structured AI outputs
- Cross-document reasoning and insights
- AI systems grounded in your specific domain model
- A reusable, long-term knowledge base for your agents
If your AI needs to understand, not just retrieve, Perseus provides the missing layer.