Text Transformation

Context

Most of an organization's knowledge is locked in unstructured text — product descriptions, requirements, reports, documentation — and pulling it into a structured knowledge graph is cumbersome, error-prone, and usually reserved for experts. Existing tools mostly just visualize whatever an AI produces, with no way to check or fix it. I built a user-centered, neuro-symbolic, human-in-the-loop tool that puts the person in charge of accuracy. Import text and the system drafts a knowledge graph; every generated statement is shown beside the text and linked by colour and connecting lines back to the exact sentence it came from, so any fact can be verified at a glance. Users explore the graph visually and edit or add statements directly, while a meaning-based confidence score shows how well each item aligns with the chosen vocabulary, and optional ontology integration keeps the output consistent with existing data models. The result is reach and trust at once: people with no prior experience produced knowledge graphs matching expert quality, in a fraction of the time — and because every fact traces back to its source, they could verify and stand behind the result. It lowers the barrier to knowledge-graph creation sharply while keeping a human firmly in control of what is true.