# Corpora ## Live today Lookup data answers exact questions, so it has no ranking benchmark. ### IP & prior art 97,913-query public holdout · trained on 609,723 examiner citations · fingerprint 6fd3de94f3b9 · [Benchmark](https://antecedint.com/docs/benchmarks) - Patents and applications10.5M+ US grants and published applications, parsed to claims and specification. - Citations48.5M examiner and applicant references, followed forward and backward. - Patent spaceEvery filing embedded and placed, so neighbours and gaps are coordinates rather than keywords. - Claim structureAntecedent basis, and which passage of the specification supports each limitation. ### Prosecution - CPC schemeThe full classification hierarchy, symbol to definition to children. - ClassificationMain CPC symbol for 8.48M publications, resolvable from an application number. - Art unitsWhere applications carrying a symbol were routed, and what each unit handles. - Examiners10,308 profiles, with how each one has read claim language before. ### Examination guidance - MPEPSections parsed to the rule each one turns on, cross-referenced both directions. - Form paragraphsThe text an examiner pastes into an office action, keyed to the rejection it carries. - EPO GuidelinesThe European counterpart, parsed the same way so the two can be read against each other. - Design codesThe design classification scheme, symbol to definition. ## Next, and underway Not searchable yet. Each is a source where someone already decided the answer. ### PTAB Decided by The Board Final written decisions name the art a petition stood or fell on. Being measured now. ### Case law, cite-check Decided by Courts, in citations and quotations Ingesting from CourtListener and GovInfo. Nothing served yet. ### Case law, authority Decided by Judges quoting precedent Which precedent a court reached for. Built on the same foundation. ### Codified law Decided by OLRC and eCFR versions The US Code and the CFR as they read on a given date. ## How a corpus becomes live 1. 1RawThe record as published 2. 2ResolvedEvery reference tied to the text it cites 3. 3JudgedThe decisions the record already holds 4. 4TunedRetrieval trained on them, eval attached The [method](https://antecedint.com/method) explains each step; the [methodology](https://antecedint.com/docs/methodology) sets the bar a number clears before it appears here. Once a corpus is live, the people who search it keep it correct: an accepted report through the MCP Feedback program goes into the next build.