Multi-Channel Rasterized Content for Document Structure Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for identifying structural elements in electronic documents, such as OCR techniques and scanning physical documents, are limited in automatically detecting structural elements, leading to errors in classification and location, especially for documents without intrinsic data, making manual input costly and inconvenient for large documents.
Innovation Solution
A methodology using a machine learning model that generates multiple channels of rasterized content, including text, graphic, and image structures, and inputting these channels to improve the identification and classification of structural elements, with additional pages before and after the target page also being rasterized and fed into the model to enhance accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional OCR techniques and scanning methods are used to identify structural elements in electronic documents, then the process is simple and straightforward, but the accuracy of detecting structural elements is limited and errors occur in classification and location
Solution Approach 1:
The patent segments the document processing task into multiple specialized components: an OCR module for text recognition, a structural element detection module for identifying document structures, and a classification module for categorizing elements. Each module focuses on specific aspects, improving overall detection accuracy while maintaining manageable system complexity through modular architecture.
Solution Approach 2:
The patent introduces an intermediary processing layer that receives OCR output and structural element detection results, then integrates and refines this information through machine learning models. This intermediary layer reconciles the outputs from different modules, improving classification accuracy without requiring complete system redesign.
2Measurement precision
If manual input is used to correct classification and location of structural elements, then accuracy can be improved, but the cost and time consumption increase significantly for large documents
Solution Approach 1:
The patent implements a self-service mechanism where the system automatically detects, classifies, and corrects structural element errors using machine learning models trained on document structures. The system performs self-validation and self-correction, eliminating the need for manual review and significantly reducing time consumption while maintaining high accuracy.
Solution Approach 2:
The patent incorporates feedback loops where the system continuously learns from detected structural elements and classification results. Machine learning models are trained on identified patterns and automatically adjust classification rules, enabling the system to improve accuracy over time without manual intervention and automatically correct previously identified errors.
3Measurement precision
If single-channel rasterized content is used as input to machine learning models, then the processing is computationally efficient, but the ability to accurately identify different types of structural elements is limited
Solution Approach 1:
The patent segments the visual document content into multiple specialized channels: text channels for OCR processing, image channels for graphical element detection, and structural channels for layout analysis. Each channel processes specific types of content with optimized algorithms, improving identification accuracy for different element types while managing computational resources through targeted processing rather than uniform analysis of all content.
Solution Approach 2:
The patent applies partial processing by focusing computational resources on specific channels based on document characteristics. Rather than processing all channels equally, the system selectively intensifies processing for channels containing relevant structural elements, achieving high accuracy for critical elements while conserving computational resources on less important content areas.
Data Source
AI summary
Techniques are provided for identifying structural elements of a document. One Methodology includes generating a first channel of rasterized content by rasterizing a full page of the document and generating one or more additional channels of rasterized content from the page of the document by rasterizing one or more corresponding content types from the page of the document. Each of the one or more additional channels includes a specific type of content that is different from each of the other one or more additional channels. The methodology further includes inputting the first channel of rasterized content and the one or more additional channels of rasterized content into a machine learning (ML) model. The methodology continues with determining location and classification for each of a plurality of structural elements on the page of the document using the ML model.