Cross-Region Document Content Recognition Through Tail-Head Merging
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for recognizing document content in untagged documents fail to recognize continuous content across different pages and columns, requiring separate operations on each page and column, making unified operations impossible.
Innovation Solution
A cross-region document content recognition method that performs document structure recognition on untagged documents, identifies tail and head document structure results based on preset area conditions, and merges these results if they are compatible, creating a unified document structure recognition set.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If document content recognition is performed on a single document block in a single page, then the recognition process is simple and fast, but continuous document content in different pages and columns cannot be recognized
Solution Approach 1:
The patent merges document structure recognition results across different pages and columns by identifying continuous content relationships. The system combines tail region results from one page/column with head region results from the next page/column when they form continuous content, creating a unified recognition result set that spans multiple regions while maintaining recognition efficiency.
2Ease of operation
If separate operations are performed on each page and column, then the operation process is simple, but unified operations on continuous document content become impossible
Solution Approach 1:
The patent introduces a document structure recognition result set as an intermediary data structure that stores and manages recognition results across multiple pages and columns. This intermediary structure enables the system to maintain relationships between continuous content elements while providing a unified interface for operations, allowing users to perform single operations on content spanning multiple regions.
3Adaptability or versatility
If document structure recognition results are merged across pages and columns, then unified operations become possible, but the complexity of the recognition process increases
Solution Approach 1:
The patent segments the document into distinct regions (head region and tail region) on each page/column and processes them separately before merging. This segmentation allows the system to manage complexity by handling small, manageable units independently while maintaining the ability to combine them into comprehensive cross-region recognition results through structured merging operations.
Data Source
AI summary
Embodiments of the present disclosure disclose a cross-region document content recognition method, device, apparatus, medium, and program product. A specific implementation of the method includes: performing a document structure recognition processing on an untagged document; for every two adjacent document page numbers in the document page number sequence: selecting a document structure recognition result that corresponds to a first document page number and satisfies a preset tail area condition, as a tail document structure recognition result; selecting a document structure recognition result that corresponds to a second document page number and satisfies a preset head area condition, as a head document structure recognition result; performing a merged detection on the tail document structure recognition result and the head document structure recognition result; in response to the tail document structure recognition result and the head document structure recognition result being mergeable.