How it works
VERGA explores context-relevant requirements retrieval and validation from technical documents with Thales, Beslogic and Confiance IA. The supplier pipeline uses Docling/Unstructured extraction, applicability domains, chapters and document coordinates before vector indexing. An Angular/AG-Grid interface combines tag/keyword/applicability filters with natural-language semantic retrieval, preserving the original PDF page and context. Confiance IA independently identifies the requirements/specification/test-link research and lists the project as ongoing. Human expertise remains central; no accepted naval product or measured retrieval benchmark is established.
Technical detail
Core features
- Docling/Unstructured extraction with applicability-domain, chapter and positional metadata