docflow-invoice-parser-fa

DocFlow Invoice Parser FA

Enterprise-grade Persian invoice field extraction model configuration by Aria AI.

Model Description

DocFlow AI combines vision OCR with NLP heuristics and business-rule validation to extract structured data from Persian and bilingual financial documents. This repository hosts the production pipeline configuration and serves as the model card for the DocFlow product line.

Intended Uses

• Accounts payable automation
• Invoice digitization for ERP integration (SAP, Rahkaran, Sepidar)
• Pilot deployments with accountant-in-the-loop review

Out-of-Scope Uses

• Legal contract analysis (see ContractGuard)
• Processing real PII in public Spaces without on-prem deployment
• Fully autonomous posting without human review

Training Data

Synthetic and anonymized invoices from docflow-invoice-samples-fa.

Evaluation Metrics (Internal Pilot)

Limitations

• Heuristic demo pipeline; accuracy varies on handwritten or low-quality scans
• Jalali date parsing does not validate calendar correctness
• Line-item table detection requires structured tabular layout

Bias & Fairness

Trained primarily on standard printed Persian invoices from manufacturing and distribution sectors. Performance may be lower on handwritten or non-standard layouts.

Demo

Interactive Space: docflow-invoice-parser-fa

Citation

@misc{docflow_invoice_parser_fa_2025,
  title={DocFlow AI: Intelligent Persian Invoice Parser},
  author={Aria AI Engineering Team},
  year={2025},
  url={https://huggingface.co/alirezaaminzadeh/docflow-invoice-parser-fa}
}

License

Apache 2.0