Semiont: A Graph-Based, AI-Native Wiki and Annotator

Session Track: Knowledge Graphs & GraphRAG

Session Time: 01:00 pm - 01:30 pm April 15

Session description

Modern AI systems increasingly centralize meaning inside proprietary embeddings and cloud-hosted black boxes—a form of “knowledge compression” where structure is stored inside the model rather than in user-controlled infrastructure. Semiont takes the opposite approach. It is an open, AI-native annotation and semantic layer designed to help communities reconstruct durable, verifiable knowledge under their own governance.

Built on the W3C Web Annotation standard, Semiont creates a rich lexical graph from documents and media—capturing references, relationships, and contextual signals that form the substrate for a transparent, ontology-informed semantic layer. This structure allows users and agents to organize meaning in a way that is inspectable, exportable, and fully under local control. Annotations also provide a practical bridge between real-world workflows and structured knowledge, enabling review processes, follow-up questions, and risk identification to coexist with the underlying content.

LLMs assist in detection, suggestion, and drafting, but Semiont is the site of reconstruction, where AI-generated insights become persistent, auditable knowledge aligned with the community’s needs rather than trapped in proprietary systems. This talk introduces Semiont’s architecture and shows how it supports privacy-preserving, community-hosted knowledge bases and lays the groundwork for decentralized or federated knowledge networks. Semiont is still early in its development, and we welcome design feedback and community participation as the architecture continues to evolve.

The project is fully open-source and available at github.com/The-AI-Alliance/semiont