Structured Document Annotation Using Template-Based Filling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for automatic annotation of structured documents, such as invoices and receipts, are inefficient due to the need for manual labor and random word placement, which fails to ensure correct content and position alignment with structured data formats.
Innovation Solution
A method that generates target filling information based on attribute values, historical content, and positions within a template image to create annotated structured documents, allowing for rapid and accurate annotation by adjusting content and position while maintaining semantic consistency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual annotation is used for structured documents, then annotation accuracy and correctness can be ensured, but labor cost and time consumption increase significantly
Solution Approach 1:
The patent uses template images as copies of structured document layouts and fills them with generated content to create annotated documents. This copying approach preserves the structural accuracy of templates while enabling rapid generation of diverse annotated examples without manual re-annotation, thus maintaining annotation accuracy while significantly reducing time consumption.
Solution Approach 2:
The patent performs preliminary actions by pre-processing template images and pre-defining field positions before generating annotated documents. The template images are prepared in advance with their structural information, and the annotation generation process simply involves filling predefined fields rather than performing complex annotation tasks from scratch, thereby reducing annotation time while maintaining precision.
2Ease of manufacture
If random word placement method is used for automatic annotation, then annotation process is simplified, but content and position alignment with structured data formats becomes incorrect
Solution Approach 1:
The patent applies local quality by treating different regions of the template image differently - specific field positions are identified and marked for content generation, while other regions remain as template structure. This localized approach ensures that content is generated only where needed and maintains proper alignment with structured data formats, avoiding the random placement problem while keeping the process simple.
Solution Approach 2:
The patent introduces template images as intermediaries between the simple generation process and the requirement for precise content-position alignment. The templates serve as mediators that encode the correct structural information, allowing the generation process to focus only on filling content without worrying about positioning, thus achieving both simplicity and precision.
3Extent of automation
If existing automatic annotation methods are used for structured documents, then manual labor is reduced, but the methods cannot handle documents with complex layouts and variations
Solution Approach 1:
The patent achieves universality by using template images that can represent multiple variations of structured document layouts. A single template system can handle different document types and layout variations by simply changing the template image, making the automatic annotation method adaptable to complex layouts without requiring separate processing methods for each document type.
Solution Approach 2:
The patent applies parameter changes by adjusting the template image parameters (such as field positions, formats, and structures) to match different document layouts. This allows the same automatic annotation system to adapt to various complex layouts by modifying template parameters rather than changing the fundamental generation process, thereby achieving both automation and adaptability.
Data Source
AI summary
Disclosed are a method, apparatus and electronic device for annotating information of a structured document. A specific implementation is: obtaining a template image of a structured document and at least one piece of annotation information of a field to be filled in the template image, where the annotation information includes attribute value and historical content of the field to be filled, and historical position of the field to be filled in the template image; generating, according to the attribute value of the field to be filled, the historical content of the field to be filled and the historical position of the field to be filled in the template image, target filling information of the field to be filled; obtaining, according to the target filling information of the field to be filled, an image of an annotated structured document.