Custom Models for Your Documents
What are Custom Models?
Custom models are machine learning models trained specifically on your organization's documents. They combine the power of optical character recognition (OCR) with intelligent field extraction, learning the patterns and structures unique to your document types.
Unlike pre-built models that work with generic documents, custom models learn from your labeled data to achieve superior accuracy on your unique document formats.
Key Capabilities
Document-Specific Training
Train models on your exact document types for unmatched accuracy.
Field-Level Customization
Define and extract custom fields unique to your business needs.
Table Recognition
Train models to recognize complex table structures effortlessly.
Multi-Language Support
Build models for documents in any language worldwide.
Continuous Improvement
Retrain models as your documents and needs evolve.
High Performance
Process thousands of documents per hour with consistent quality.
Pre-Built vs Custom Models
| Feature | Pre-Built Models | Custom Models |
|---|---|---|
| Setup Time | Immediate | Requires training data |
| Accuracy on Standard Docs | High | High |
| Accuracy on Custom Docs | Medium | Very High |
| Custom Fields | Limited | Unlimited |
| Industry-Specific | No | Yes |
| Adaptation | Fixed | Continuous learning |
Benefits of Custom Models
Superior Accuracy: Custom models achieve 15-30% higher accuracy on organization-specific documents compared to generic models, reducing extraction errors and manual review time.
Flexible Field Definitions: Define any field type relevant to your business including custom invoice fields, industry-specific identifiers, proprietary document sections, and multi-value patterns.
Reduced Manual Work: Higher accuracy means fewer documents requiring manual review, faster processing times, lower operational costs, and improved employee satisfaction.
Scalability
- ✓ Handle thousands of documents per hour
- ✓ Consistent quality across all documents
- ✓ Automatic processing without human intervention
Competitive Advantage
- ✓ Faster document turnaround times
- ✓ Better customer experience
- ✓ Data-driven decision making
- ✓ Reduced processing costs
Industry Use Cases
Financial Services
Invoice Processing
- • Vendor-specific invoice formats
- • Custom line item structures
- • Multi-currency handling
- • Tax calculations and breakdowns
Loan Applications
- • Bank statements with varied formats
- • Employment verification documents
- • Tax returns and financial statements
Healthcare
Medical Records
- • Clinical notes with medical terminology
- • Lab reports with complex tables
- • Prescription forms
- • Insurance claim forms
Patient Intake
- • Registration forms
- • Insurance cards
- • Medical history questionnaires
Legal
Contract Analysis
- • Parties and signatories
- • Key dates and deadlines
- • Financial terms and conditions
- • Clauses and obligations
Discovery Documents
- • Email threads with metadata
- • Legal briefs and filings
- • Evidence documentation
Logistics & Supply Chain
Shipping Documents
- • Bills of lading with carrier formats
- • Packing lists with item structures
- • Customs declarations
- • Delivery receipts
Purchase Orders
- • Multi-supplier PO formats
- • Line items with specifications
- • Payment terms
Getting Started: Prerequisites
What You Need
- ✓ Sample Documents: 50-200 representative documents
- ✓ Field Definitions: Clear list of fields to extract
- ✓ User Permissions: Model training access in DMind
- ✓ Time Allocation: 2-4 hours for initial labeling
Document Requirements
Quality Guidelines:
- • Resolution: Minimum 150 DPI (300 DPI recommended)
- • Format: PDF, PNG, JPEG, TIFF
- • File Size: Up to 50MB per document
- • Clarity: Clear, legible text
Diversity Requirements:
- • Include documents from different time periods
- • Represent various vendors/sources
- • Include edge cases and variations
- • Mix of good and challenging quality
