Form Document Processing Through Spatial Entity Relationships
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Computing systems struggle to accurately process form documents with varying formats, as existing optical character recognition (OCR) solutions only retrieve text sequentially and fail to capture multi-directional and indirect relationships between entities.
Innovation Solution
A computing system identifies entities in a form document, generates vectors describing content and position, applies a classification machine-learned (ML) model to determine semantic relationships between entities using a Siamese neural network, and processes the form document based on these relationships.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If sequential text retrieval methods are used, then the processing simplicity is maintained, but the ability to capture multi-directional relationships between entities deteriorates
Solution Approach 1:
The patent transitions from sequential one-dimensional text processing to two-dimensional spatial relationship analysis by generating bounding boxes and calculating spatial distances between entities. This dimensional expansion enables the system to capture multi-directional relationships (above, below, left, right) that sequential methods cannot detect, directly resolving the contradiction between processing simplicity and relationship information capture.
2Measurement precision
If format-specific processing rules are created for different form layouts, then the processing accuracy for specific formats is improved, but the system complexity increases
Solution Approach 1:
The patent creates a universal processing framework that handles diverse form formats through a single spatial relationship-based approach. Instead of developing separate processing rules for different form layouts, the system uses bounding box coordinates and spatial distance calculations that work across all formats, thereby maintaining high processing accuracy while avoiding the complexity of multiple format-specific systems.
Solution Approach 2:
The patent changes the processing parameters from format-specific text patterns to universal spatial coordinates and distance metrics. By representing entities as bounding boxes with coordinates (x1, y1, x2, y2) and calculating spatial relationships based on these parameters, the system achieves format-agnostic processing that maintains accuracy across different layouts without requiring complex format-specific rules.
3Ease of manufacture
If traditional OCR text extraction is used, then the implementation simplicity is maintained, but the understanding of semantic relationships between entities deteriorates
Solution Approach 1:
The patent segments the form document into discrete entities with bounding boxes, allowing independent identification and relationship analysis of each entity. This segmentation enables the system to maintain implementation simplicity through automated coordinate-based processing while significantly improving semantic relationship understanding by explicitly calculating spatial relationships between segmented entities rather than treating text as a continuous stream.
Data Source
AI summary
A method comprises identifying entities in a form document; for each of the entities, generating a vector for the entity that comprises a plurality of values describing content of the entity and a position of the entity within the form document; identifying, by the computing system, pairs of the entities that satisfy a spatial relationship requirement; for each pair of entities satisfying the spatial relationship requirement, applying a classification machine-learned (ML) model that takes as input the vectors for the pair of entities and outputs a category for a first entity of the pair of entities and a category for a second entity of the pair of entities, wherein a combination of the category for the first entity and the category for the second entity represents a semantic relationship between the first entity and the second entity; and processing the form document based on the semantic relationships between the entities.