A county assessor needs deed-related scans searchable by extracted fields after OCR. Which pipeline combination should they run?

A county assessor making deed-related scans searchable by extracted fields should run OCR combined with field extraction for indexing. OCR converts deed scan pixels to text; field extraction promotes parties, parcel identifiers, recording numbers, and dates into indexable fields staff can filter and search. OCR with full-text indexing on deed-related scans for keyword retrieval turns pages into searchable text but does not promote grantor, grantee, parcel, or recording fields into structured filters assessors can query. OCR with field extraction on deed-related scan metadata for indexing reads the pages but only promotes scan file metadata into the index, missing the parties, parcel identifiers, recording numbers, and dates trapped in the document content itself. OCR with document classification on deed-related scans for indexing labels scans as deeds for routing but does not promote grantor, grantee, parcel, or recording fields into structured filters, so the scans remain unsearchable by grantor, grantee, or parcel ID. Exam caveat: searchable municipal deed scans require OCR plus field extraction into the index—not full-text indexing alone, field extraction on scan metadata alone, or document classification alone. Operational check: OCR deed scans; extract recording and parcel fields with confidence scores; load fields into the assessor search index; test queries by grantor, grantee, and parcel ID against known deeds before retiring document-classification-only access for those attributes. Store page-level OCR confidence with each extracted deed field so low-confidence party or parcel values route to examiner review instead of silent index pollution.