Table Structure Preservation in Automatic Text Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing text recognition systems fail to maintain the structure of tables when copying or transferring text data from images, resulting in unstructured bulk text that lacks context, making manual transcription time-consuming and error-prone.
Innovation Solution
A machine learning approach that recognizes, extracts, and reconstructs tables from images, using a trained model to identify text and table structures, allowing for direct copying and pasting while preserving the original table format.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If traditional OCR systems are used to recognize text from images, then text recognition is achieved, but the table structure is lost and unstructured bulk text is produced
Solution Approach 1:
The patent segments the text recognition process into distinct components: table structure detection, cell boundary identification, and text extraction within cells. This segmentation allows the system to preserve the hierarchical structure of tables while extracting text, preventing the loss of organizational information that occurs in traditional OCR approaches.
Solution Approach 2:
The patent introduces an intermediary table structure model that acts as a mediator between the image input and the extracted text output. This intermediary representation maintains the table's structural information (rows, columns, cells) while enabling text extraction, thus preventing direct loss of structure that would occur in conventional OCR systems.
2Reliability
If manual transcription of tables is performed to preserve structure, then table structure is maintained, but time consumption and error rate increase
Solution Approach 1:
The patent enables the system to automatically perform the task that previously required manual intervention. The table structure recognition and text extraction are performed autonomously by the system, eliminating the need for manual transcription while maintaining structural integrity. The system serves itself by automatically identifying and preserving table structures without human assistance.
Solution Approach 2:
The patent replaces the mechanical manual transcription process with an automated computational system. Instead of human operators manually copying table structures, the system uses algorithmic approaches to detect, recognize, and preserve table structures automatically, substituting human mechanical action with computational processing.
3Loss of information
If traditional text recognition methods are used, then text extraction is achieved, but context and structure are lost
Solution Approach 1:
The patent adds a structural dimension to the text extraction process. Instead of extracting only flat text content, the system extracts text with its positional and hierarchical context preserved through table structure information. This dimensional enhancement maintains the relationship between text elements and their structural context.
Solution Approach 2:
The patent performs preliminary detection and recognition of table structures before text extraction occurs. By identifying the table structure in advance, the system prepares the contextual framework that will guide subsequent text extraction, ensuring that text is extracted with its proper structural context intact rather than as isolated content.
Data Source
AI summary
Aspects of the subject technology include identifying one or more portions of a data object that include a table by providing the data object to a table detection model, determining a structure of the table by providing the one or more portions of the data object to a table structure recognition model, generating a virtual table based on the determined structure of the table, the virtual table including an indication of at least one of one or more rows, one or more columns, or one or more cells corresponding to the table, mapping text from the one or more portions of the data object to corresponding cells of the virtual table, and performing a process with the virtual table.