How the registry is built
I build the registry from information that is already publicly available online, then organize, compare and connect it into a more useful picture.
Sources
Sources include company websites, patent databases, scientific papers, conference and exhibition materials, university and research-institute pages, technical publications and public directories.
Identity and merge rules
Records are not merged by similarity alone. Publication identity is consolidated by exact DOI or an exact alternate identifier such as a conference or arXiv identifier. Ambiguous records remain separate until exact evidence is available. Fuzzy matching is not used for canonical merges.
Data layers
Physical preserves source and historical rows. Controlled contains verified records accepted into the working registry. Canonical is the normalized identity layer used for analysis. Physical evidence-only preserves valid provenance that is not allowed to inflate controlled KPIs. Public is the selected aggregate projection shown on this site.
Current counting rules
Company counts distinguish 2,302 normalized company entities from 7,115 controlled company records. Patent-family counts distinguish 1,124 strict/public families, 1,191 controlled families and 23 physical evidence-only nodes; the visible graph therefore contains 1,214 patent-family nodes. Publication counts distinguish 2,706 physical publication rows from 2,620 active canonical publications. The synchronized bibliography contains 2,324 unique publications and 3,392 Publication ↔ Topic edges.
Machine-readable metric definitions and current values
Relations and derived analytics
Direct evidence relations and analytically derived relations are stored separately. Derived topic or cross-layer mappings do not automatically become canonical evidence.
Integrity
The registry is append/consolidate-only: accepted IDs and evidence are not silently deleted or recycled. Registry Core v2.5 FINAL passed SQLite integrity checks, has zero foreign-key violations, a 14/14 reverse-loss audit and zero canonical losses.
Independent status
The registry is independently developed and maintained. Being listed does not mean that a company or organization endorses the registry or is formally connected with it.
Corrections
If a public aggregate or published organization record is incorrect, please contact me with the source that supports the correction.