Segment review

Every title package alongside the segments a document was split into. Open one to read the segment list against the pages themselves.

46 documents 562 segments2,390 pages 5 duplicates removed 18 to review first

Labelled set

title-labelled-docs

Six packages whose delivered PDFs carried an index sheet listing every document and its page range. Segments here are the ground truth, corrected against page content.

Title Review C
60 pages·16 segments·index-sheet ground truth
Review first14.8
Lien ×4Deed ×3Probate Document ×2Summary Tax Other Divorce Decree Affidavit Mortgage Assignment
complex doc types: divorce decree, probate document, lien, affidavit · 4 page(s) unidentified (7% of the package) · 10 distinct doc types, 4 beyond the routine set
Title Review D
59 pages·12 segments·index-sheet ground truth
Review first13.1
Deed ×2Other ×2PACER Search ×2Summary Tax Mortgage Assignment Probate Document Divorce Docket / Register of Actions
complex doc types: divorce docket, probate document, pacer search · 2 page(s) unidentified (3% of the package) · 9 distinct doc types, 3 beyond the routine set
Title Review B
76 pages·21 segments·index-sheet ground truth
Review first8.3
Deed ×9Mortgage ×4Affidavit ×3Other ×2Summary Assignment Death Certificate
complex doc types: death certificate, affidavit · 6 page(s) unidentified (8% of the package) · 7 distinct doc types, 2 beyond the routine set
Title Review F
26 pages·7 segments·index-sheet ground truth
Worth a look6.0
Deed ×2Mortgage ×2Summary Death Certificate Lien
complex doc types: death certificate, lien · 5 distinct doc types, 2 beyond the routine set
Title Review E
39 pages·8 segments·index-sheet ground truth
Worth a look4.0
Mortgage ×3Deed ×2Summary Assignment Divorce Docket / Register of Actions
complex doc types: divorce docket · 5 distinct doc types, 1 beyond the routine set
Title Review A
49 pages·14 segments·index-sheet ground truth
Routine1.0
Deed ×6Mortgage ×3Assignment ×3Summary Tax
5 distinct doc types

Eyal set

eyal-docs

Fourteen held-out packages with no labels. Segments here are the pipeline's own predictions (gemini-3.5-flash, general profile) — this is what needs a human verdict.

New docs

new-docs

Twenty-six Kansas, Nebraska and Iowa example packages delivered as zip archives. Parsed and segmented by the same pipeline (gemini-3.5-flash, general profile); no labels and no review pass yet.

Deduplicated (5)

Title Report 12.31.24 - Marked Up - 250735 — annotated copy of a package already in the list - not parsed or segmented (see eyal-docs/PROVENANCE.md)
Title Report 1_20_26 - Marked Up - 260483 — annotated copy of a package already in the list - not parsed or segmented (see eyal-docs/PROVENANCE.md)
Title Report 4_1_26 - Marked up - 260640 — annotated copy of a package already in the list - not parsed or segmented (see eyal-docs/PROVENANCE.md)
Title Report 6_9_26 - marked up - 264477 — annotated copy of a package already in the list - not parsed or segmented (see eyal-docs/PROVENANCE.md)
Title Report 7_22_25 - Marked up - 260757 — annotated copy of a package already in the list - not parsed or segmented (see eyal-docs/PROVENANCE.md)