{ "synthetic": true, "description": "Ground-truth key fields for the workshop OCR companion docs. Each document has 5 tier variants: clean (baseline) and tier1..tier4 (increasing degradation). Participants attempt to extract these fields from the tier N PDFs; accuracy is scored against this file.", "documents": [ { "doc_id": "nsf_award_OAC-2415678", "doc_type": "NSF Award Notice", "sponsor": "NSF", "pages": 4, "clean_pdf": "workshop_ocr/clean/nsf_award_notice_OAC-2415678.pdf", "tiered_pdfs": { "tier1": "workshop_ocr/tier1_structural/nsf_award_notice_OAC-2415678.pdf", "tier2": "workshop_ocr/tier2_scan/nsf_award_notice_OAC-2415678.pdf", "tier3": "workshop_ocr/tier3_authentic/nsf_award_notice_OAC-2415678.pdf", "tier4": "workshop_ocr/tier4_layout/nsf_award_notice_OAC-2415678.pdf" }, "ground_truth_fields": { "award_number_fain": "OAC-2415678", "amendment_number": "000", "award_instrument": "Continuing Grant", "award_date": "2024-08-05", "period_of_performance_start": "2024-09-01", "period_of_performance_end": "2028-08-31", "project_title": "Trustworthy campus-scale assistants for research lifecycle intelligence", "funding_opportunity": "NSF 24-512 Cyberinfrastructure for Sustained Scientific Innovation (CSSI)", "assistance_listing": "47.070", "amount_obligated_this_amendment": 462500.00, "total_intended_award_amount": 1850000.00, "cost_share_amount": 0.00, "principal_investigator": "Malik T. Ashford", "pi_email": "mashford@synthetic.uidaho.edu", "pi_organization": "UNIVERSITY OF IDAHO", "recipient_legal_name": "REGENTS OF THE UNIVERSITY OF IDAHO (SYNTHETIC)", "recipient_uei": "SYN7QWYKRJH5", "recipient_address": "875 PERIMETER DR MOSCOW, ID 83844-9803", "managing_program_officer": "Dr. Amelia K. Voss", "managing_program_officer_email": "avoss@synthetic.nsf.gov", "managing_grants_official": "Priya R. Lakshmi", "awarding_official": "Jannele L. Gosey", "indirect_cost_rate_pct": 50.0, "indirect_cost_base": "Modified Total Direct Costs", "budget": { "total_salaries_and_wages": 1028000.00, "fringe_benefits": 308400.00, "equipment": 0.00, "domestic_travel": 49333.00, "international_travel": 0.00, "materials_supplies": 61667.00, "publication_costs": 8000.00, "computer_services": 24667.00, "total_direct_costs": 1233334.00, "indirect_costs": 616666.00, "total_direct_and_indirect": 1850000.00 } } }, { "doc_id": "subaward_SUB-24-001", "doc_type": "Subaward Agreement", "sponsor_upstream": "NIH (via University of Idaho prime recipient)", "pages": 3, "clean_pdf": "workshop_ocr/clean/subaward_agreement_SUB-24-001.pdf", "tiered_pdfs": { "tier1": "workshop_ocr/tier1_structural/subaward_agreement_SUB-24-001.pdf", "tier2": "workshop_ocr/tier2_scan/subaward_agreement_SUB-24-001.pdf", "tier3": "workshop_ocr/tier3_authentic/subaward_agreement_SUB-24-001.pdf", "tier4": "workshop_ocr/tier4_layout/subaward_agreement_SUB-24-001.pdf" }, "ground_truth_fields": { "subaward_number": "SUB-24-001", "prime_award_number": "R01-AI-24-8002", "prime_sponsor": "NIH", "subaward_type": "Cost-Reimbursement Subaward", "executed_date": "2024-08-15", "period_of_performance_start": "2024-09-01", "period_of_performance_end": "2028-08-31", "total_obligated_amount": 185000.00, "f_and_a_rate_pct": 26.0, "f_and_a_type": "off_campus", "f_and_a_base": "MTDC", "prime_cfda": "93.855", "pass_through_entity": "Regents of the University of Idaho (Synthetic)", "pass_through_uei": "SYN7QWYKRJH5", "prime_pi": "R.J. MacReady, Ph.D.", "subrecipient_organization": "Outpost 31 Arctic Research Station", "subrecipient_uei": "O31ARST0031X", "subrecipient_ein": "99-0310310", "subrecipient_cage": "O31RS", "subrecipient_duns": "031031031031", "subrecipient_authorized_official": "Gary Bennings, Station Director", "subrecipient_pi": "Blair A. Copper", "subrecipient_pi_email": "bcopper@outpost31.org", "budget_periods": [ {"period": 1, "start": "2024-09-01", "end": "2025-08-31", "amount": 46250.00}, {"period": 2, "start": "2025-09-01", "end": "2026-08-31", "amount": 46250.00}, {"period": 3, "start": "2026-09-01", "end": "2027-08-31", "amount": 46250.00}, {"period": 4, "start": "2027-09-01", "end": "2028-08-31", "amount": 46250.00} ], "overall_risk_rating": "Medium", "risk_factor_geographic": "High", "risk_assessed_by": "Palmer, Nauls K. (Post-Award Financial Analyst)", "risk_assessment_date": "2024-07-20", "invoices": [ {"invoice_number": "O31-INV-2025-001", "amount": 28750.00, "status": "Paid"}, {"invoice_number": "O31-INV-2025-002", "amount": 17500.00, "status": "Under Review"} ] } }, { "doc_id": "nih_noa_1R01AI248002-01", "doc_type": "NIH Notice of Award", "sponsor": "NIH/NIAID", "pages": 3, "clean_pdf": "workshop_ocr/clean/nih_notice_of_award_1R01AI248002-01.pdf", "tiered_pdfs": { "tier1": "workshop_ocr/tier1_structural/nih_notice_of_award_1R01AI248002-01.pdf", "tier2": "workshop_ocr/tier2_scan/nih_notice_of_award_1R01AI248002-01.pdf", "tier3": "workshop_ocr/tier3_authentic/nih_notice_of_award_1R01AI248002-01.pdf", "tier4": "workshop_ocr/tier4_layout/nih_notice_of_award_1R01AI248002-01.pdf" }, "ground_truth_fields": { "sponsor_award_number": "1R01AI248002-01", "fain": "1R01AI248002", "federal_award_date": "2024-09-01", "award_type": "New", "action_type": "New Competing Award", "budget_period_start": "2024-09-01", "budget_period_end": "2025-08-31", "project_period_start": "2024-09-01", "project_period_end": "2028-08-31", "award_title": "Molecular mechanisms of cellular identity subversion in Thingium assimilans", "nih_activity_code": "R01", "funding_opportunity": "PAR-23-218", "assistance_listing_cfda": "93.855", "assistance_listing_name": "Allergy, Immunology, and Transplantation Research", "recipient_name": "REGENTS OF THE UNIVERSITY OF IDAHO (SYNTHETIC)", "recipient_address": "875 PERIMETER DR, MOSCOW, ID 83844-9803", "recipient_uei": "SYN7QWYKRJH5", "recipient_ein": "82-0290662", "congressional_district": "02", "pi_name": "R.J. MACREADY, PhD", "pi_email": "rmacready@synthetic.uidaho.edu", "authorized_official": "PEREGRIN A. TOOKFIELD", "grants_management_specialist": "SAMANTHA P. HENEGAR", "grants_management_email": "shenegar@synthetic.niaid.nih.gov", "program_official": "DR. NICOLAS J. DEMBOWSKI", "program_official_email": "ndembowski@synthetic.niaid.nih.gov", "year1_budget": { "salaries_and_wages": 245000.00, "fringe_benefits": 73500.00, "consultant_services": 0.00, "equipment": 0.00, "supplies": 28000.00, "travel": 14500.00, "other_expenses": 11750.00, "consortium_contractual_direct": 27250.00, "total_direct_costs": 400000.00, "indirect_costs_fa_50pct_mtdc": 200000.00, "total_approved_budget": 600000.00, "federal_share": 600000.00, "non_federal_share": 0.00 }, "future_year_commitments": [ {"year": 2, "direct": 400000.00, "indirect": 200000.00, "total": 600000.00}, {"year": 3, "direct": 400000.00, "indirect": 200000.00, "total": 600000.00}, {"year": 4, "direct": 400000.00, "indirect": 200000.00, "total": 600000.00} ], "priority_score": 18, "percentile": 7, "clinical_trial_indicator": "No", "human_subjects_code": "10", "iacuc_approval_id": "UI-IACUC-2024-037", "approved_subaward_amount": 185000.00, "approved_subaward_id": "SUB-24-001" } } ], "degradation_tiers": { "clean": "Ground-truth PDF as generated. Full text layer. Naive pdftotext extraction recovers 100%.", "tier1_structural": "Image-only PDF (no text layer), 150 DPI, light skew, JPEG Q=55. Minimum change that defeats naive pdftotext — now requires actual OCR. All visual content is still crisp.", "tier2_scan": "Image-only. Multi-page bleed-through (mirrored faint text from adjacent page), Gaussian noise, salt-and-pepper speckle, mild blur, aggressive contrast reduction, 1-bit Floyd-Steinberg dithering (fax/photocopy aesthetic), horizontal banding streaks, 40-quality JPEG re-compression. Targets tesseract and generic OCR engines.", "tier3_authentic": "Builds on Tier 2. Adds red 'RECEIVED' stamp overlapping header text, yellow highlighter over budget figures (with semi-transparent multiply blend), handwritten signatures and margin annotations (Bradley Hand font + path-drawn squiggles), three-hole-punch cutouts on the left edge, coffee ring stain. Simulates the post-scan, post-review state that sponsor-generated PDFs typically arrive in.", "tier4_layout": "Image-only, Tier 1-level degradation + one page rotated 90 degrees + budget table replaced with a hand-drawn grid-line image in which ~45% of horizontal row separators are randomly omitted (creating a 'merged cell' ambiguity that defeats table-structure extraction)." } }