In the steel industry, accurate data extraction from material inspection certificates is critical for ensuring product quality and regulatory compliance. However, extracting key details such as charge, order numbers, and complex chemical composition tables is challenging, especially as each supplier uses different formats. Our solution, cbs AID (Advanced Integration of Documents), streamlines this process using artificial intelligence.
The Initial Situation
Document processing in two steps
Step 1: Document Reading
cbs AID begins with the Document Reading step, ensuring every certificate is ready to be analysed. This step includes:
- Rotation Detection & Correction: Our system automatically detects and corrects any document rotation.
- Text Conversion: Scanned images of documents are converted into text with using optical character recognition.
This phase all documents are generalized, regardless of their format, making them ready for in-depth analysis.
Step 2: Document Understanding – Tailored for Steel Inspection Certificates
After Document Reading, cbs AID proceeds with Document Understanding. Here, the system performs a contextual document analysis to comprehend and extract important data:
- Charge & Order Number Identification: Extraction of vital numerical information from a variety of layouts.
- Complex Table Understanding: Application of advanced algorithms to accurately interpret dense chemical composition tables.
A key innovation is our Area of Interest Prediction. Our algorithm identifies the most likely locations of critical data (especially the chemical composition) within the document. By focusing on these areas, accuracy increases from to an outstanding 99.79 %.
The Technology Behind cbs AID
cbs AID leverages:
- Multi-Modal Large Language Models (LLMs): These models understand both text and layout and ensure that data is interpreted in the same way as a human expert would.
- Adaptive Algorithms: Continuous learning enables our system to process diverse formats and adapt effortlessly to novel challenges.
Benefits for Clients
cbs AID delivers near-perfect extraction accuracy that minimizes errors and upholds quality control. By automating manual processes, it boosts operational efficiency and reduces costs. A single process for all suppliers simplifies maintenance and eliminates frequent retrainings, while reliable data extraction ensures regulatory compliance.
