Focused on the field of international settlement, utilizing ICR (AI-based OCR), NLP, knowledge graphs, and other AI technologies to create an intelligent bill review system.
The GienTech Intelligent Bill Review System is a system focused on international settlement, combining ICR (AI-based OCR), NLP, knowledge graphs, and other AI technologies. It enables the automatic recognition of image-based documents, semantic understanding of letter of credit terms, configurable review rules, and automatic matching and verification through a rules engine, thereby achieving automatic review of international settlement documents and generating non-compliance reports.
Interpret message terms, automatically review non-compliance points, and transition the bill review process from offline to online, greatly enhancing review efficiency. Compared to manual bill review, business efficiency is significantly improved, with document recognition efficiency increased by 100 times and review efficiency increased by 300 times.
- ICR (Intelligent Character Recognition)
ICR technology is used to correctly recognize and extract image data, outputting the document elements needed for intelligent bill review.
- NLP (Natural Language Processing)
NLP technology is used to interpret letter of credit terms and extract structured data, which serves as the basis for review.
- Knowledge Graph
Knowledge graphs are used to build a networked knowledge system, combined with the review rules engine to achieve automatic document review.
- Rules Engine
The rules engine automatically matches and verifies documents, enabling automatic document review.
- Deep Recognition
The intelligent bill review system focuses on ICR recognition of financial industry documents, distinguishing it from general-purpose OCR recognition. It deeply develops and adapts document recognition scenarios, making it compatible with both intelligent bill review and intelligent remittance business scenarios.
- High Accuracy
The intelligent bill review ICR achieves an accuracy of over 95% for document recognition and information extraction based on template-marked documents. For documents without template markings, the recognition and information extraction accuracy is above 80%.