Hierarchical Document Object Extraction for RPA Table Parsing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing robotic process automation (RPA) systems struggle to accurately extract data from document images, particularly from tables and key-value pairs, limiting their reliability and flexibility in processing business transaction documents.
Innovation Solution
A hierarchical data extraction method that detects and categorizes objects within documents at multiple levels, using object detection and classification to direct different types of objects to specific data extraction components, enabling precise and reliable extraction of key-value pairs and table data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If OCR processing is used to extract text from document images, then text recognition is achieved, but data extraction from tables and key-value pairs is not well suited
Solution Approach 1:
The patent segments the document processing task into multiple hierarchical levels: first detecting individual objects (text lines, tables, key-value pairs), then categorizing them by type, and finally extracting data using specialized components for each object type. This segmentation allows OCR to handle text recognition while separate specialized components handle table and key-value pair extraction, resolving the contradiction between text recognition accuracy and data extraction versatility.
Solution Approach 2:
The patent creates a universal data extraction system that handles multiple document types and structures through a single integrated platform. The system uses an object classifier to direct different object types to appropriate extraction components, enabling one system to universally extract data from tables, key-value pairs, and other document elements, thereby achieving both text recognition and versatile data extraction capabilities.
2Device complexity
If a single-level extraction process is used, then processing simplicity is maintained, but extraction reliability and precision are reduced
Solution Approach 1:
The patent divides the extraction process into hierarchical levels: object detection at the first level, object classification by type at the second level, and type-specific data extraction at the third level. This multi-level segmentation improves extraction reliability by applying specialized processing to different object types while maintaining clear process structure through systematic organization of each level's functions.
Solution Approach 2:
The patent adds dimensional complexity to the extraction process by introducing hierarchical levels and object type classification. Instead of a single flat processing dimension, the system operates across multiple dimensions: detection dimension, classification dimension, and extraction dimension. This dimensional expansion enables higher reliability through specialized processing while maintaining manageability through structured organization.
3Productivity
If generic text extraction methods are used, then processing speed is maintained, but understanding of document content (key-value pairs and tables) is insufficient
Solution Approach 1:
The patent performs preliminary object detection and classification before data extraction. By first identifying and categorizing objects (tables, key-value pairs, text lines) and their positions, the system prepares the document structure in advance. This preliminary action enables faster subsequent extraction by avoiding re-analysis during the actual data extraction phase, thus maintaining processing speed while improving content understanding.
Solution Approach 2:
The patent introduces an object classifier as an intermediary component between object detection and data extraction. This intermediary categorizes detected objects by type and directs them to appropriate extraction components, ensuring that table data, key-value pairs, and text are processed by specialized handlers. This intermediary layer preserves processing efficiency while significantly improving document content understanding by matching extraction methods to object types.
Data Source
AI summary
Improved techniques to access content from documents in an automated fashion. The improved techniques permit extraction of data from documents, namely, images of documents. The extraction processing can be hierarchical, such as being performed in multiple levels (i.e., multi-leveled). At an upper level, numerous different objects within a document can be detected along with positional data for the objects and can be categorized based on a type of object. Then, at lower levels, the different objects can be processed differently depending on the type of object. As a result, data extraction from the document can be performed with greater reliability and precision.