Document AI & OCR
From scanned paper to structured data
Built on our open-source work in Vietnamese OCR and document layout analysis, we digitize documents end-to-end: detect layout, read Vietnamese text accurately, and extract key fields from financial statements, invoices, and forms into clean structured data.
Core Capabilities
Vietnamese OCR
High-accuracy Vietnamese text recognition, robust to fonts, scans, and photos.
Layout Analysis
AI-based detection of titles, paragraphs, tables, and figures to understand document structure.
Financial Document Extraction
Key-information extraction from financial statements and reports into analysis-ready datasets.
Table Extraction
Reconstruct tables from scans into spreadsheets and databases while preserving structure.
PDF & Scan Pipeline
Batch processing pipelines for large archives: deskew, denoise, OCR, extract, and index.
LLM-assisted Understanding
Combine OCR with LLMs to classify documents, resolve ambiguity, and validate extracted fields.
Problems We Solve
Financial Statement Digitization
Automatically extract balance-sheet and income-statement figures from scanned reports for credit analysis and data platforms.
How It Works
Technology Stack
Try the Interactive Demo
Vietnamese OCR Demo
Pick a sample document, run the pipeline: layout analysis, then text recognition
Run OCR to see recognized text here.
Simulated demo with prepared samples. The real pipeline (open source on GitHub) runs YOLOv7 layout analysis + VietOCR recognition.
Related Projects
Open-source optical character recognition pipeline tuned for Vietnamese text in scanned documents and photos
Key-information extraction from scanned financial statements into structured, analysis-ready data
YOLOv7-based detection of titles, paragraphs, tables, and figures for document understanding pipelines