Graph-Based Document Layout Detection for Multi-Column OCR
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
OCR algorithms struggle to discern the layout of documents with complex structures, such as two-column layouts, and fail to identify the type of document based on its layout, leading to incorrect reading orders.
Innovation Solution
A document layout system that applies OCR to identify text, generates a graph representation of the document with nodes and edges, and uses a graph convolutional network (GCN) machine learning model to determine the layout and classify text sections, providing a hierarchical organization of the document.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If OCR algorithms are used to process documents, then text recognition is achieved, but layout information and reading order cannot be correctly identified
Solution Approach 1:
The patent segments the document into multiple text blocks based on spatial positioning and visual characteristics. Each text block is independently processed and assigned to a specific layout region (e.g., left column, right column, header, footer). This segmentation enables the system to preserve layout information while maintaining text recognition accuracy, as each segment is analyzed for its positional relationships and visual properties.
Solution Approach 2:
The patent transitions from traditional one-dimensional line-based OCR processing to two-dimensional block-based processing. By detecting text blocks and analyzing their spatial coordinates, widths, heights, and relative positions, the system adds a dimensional layer that captures layout information. This dimensional change allows the system to distinguish between different layout regions and determine reading order based on spatial relationships rather than simply processing text line by line.
2Productivity
If traditional line-based OCR processing is used, then simple text extraction is achieved, but complex multi-column layouts cannot be correctly interpreted
Solution Approach 1:
The patent performs preliminary actions by first detecting and segmenting the document into text blocks before extracting text content. This preliminary spatial analysis and block classification is conducted in parallel with text recognition, so that when text extraction is completed, the layout structure is already established. This preliminary action enables accurate layout detection in complex multi-column layouts while maintaining high processing speed, as the blocking and text extraction operations overlap rather than sequence.
Solution Approach 2:
The patent creates a universal text block representation that serves multiple functions simultaneously. Each text block contains both the extracted text content and spatial information, allowing the same data structure to be used for text extraction, layout analysis, reading order determination, and document type classification. This multi-functionality eliminates the need for separate processing pipelines, improving productivity while maintaining reliability across different document types and layouts.
3Device complexity
If OCR algorithms process documents without layout analysis, then processing simplicity is maintained, but document type classification cannot be performed
Solution Approach 1:
The patent merges text recognition and layout analysis into a unified processing framework. The same text block detection and spatial analysis mechanisms used for layout identification also serve text extraction and document type classification. By combining these functions rather than implementing separate systems, the patent maintains processing simplicity while achieving versatile document type identification capabilities.
Solution Approach 2:
The patent introduces text blocks as intermediary structures that bridge text recognition and layout analysis. These text blocks serve as mediator objects that contain both textual content and spatial information, enabling the system to transition from simple text extraction to comprehensive layout analysis and document type classification without adding significant complexity. The text block acts as an intermediary data structure that facilitates multi-functional processing.
Data Source
AI summary
A document layout system for determining a layout of a document. The document layout system is configured to apply an OCR technique to identify the text in the document. The document layout system is further configured to generate a graph representation of the document, wherein the graph representation comprises a plurality of nodes and a plurality of edges that connect different ones of the plurality of nodes, wherein individual ones of the nodes correspond to different portions of the text. The document layout system is also configured to apply a graph cluster network machine learning model to the graph representation to identify a layout of different sections of the document according to respective merge inferences determined for individual ones of the plurality of edges. The document layout system is also configured to provide the layout of different sections of the document.