Table Extraction from Static PDFs Using Coordinate Grid Mapping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for extracting tabular data from documents, such as those in the municipal bonds market, often require intensive manual processing due to the use of static or locked PDF and scanned image formats, which optical character recognition techniques fail to accurately capture formatting and structural information, leading to garbled text and inefficient data conversion.
Innovation Solution
A system that identifies and extracts tabular data by converting text characters into blocks, drawing a grid over the document to define table cells, and storing the data in a structured format like JSON, allowing for conversion to formats like Excel, HTML, or XML, using a parser and extractor module to handle formatting and position information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If optical character recognition techniques are used to extract tabular data from static or locked PDF files, then text can be retrieved, but formatting information, structural identification information and table identification information are not captured accurately
Solution Approach 1:
The patent segments the document processing into distinct modules: a parser module that extracts text and formatting information, and an extractor module that identifies table structures. This segmentation allows each module to specialize in capturing specific types of information (formatting vs. structural), resolving the contradiction between retrieving text and preserving formatting/structural integrity.
Solution Approach 2:
The patent introduces an intermediary processing system between the static PDF and the final extracted data. This intermediary system uses coordinate information to map text elements to their visual positions, reconstructing table structures by mediating between the raw PDF content and the desired structured output, thereby preserving both text accuracy and formatting information.
2Loss of information
If manual copy and pasting operations are used to retrieve text from encoded files, then text can be obtained, but the process is inefficient and time-consuming
Solution Approach 1:
The system performs self-service by automatically extracting and structuring tabular data without requiring manual intervention. The parser and extractor modules work autonomously to identify tables, extract data, and output structured results, eliminating the need for manual copy-pasting while maintaining complete text retrieval and significantly improving productivity.
Solution Approach 2:
The patent replaces the mechanical manual process of copy-pasting with an automated computational system. The system uses coordinate-based mapping and pattern recognition algorithms to automatically identify and extract tabular data, substituting human manual labor with automated processing that is both faster and more accurate.
3Adaptability or versatility
If intensive pre-processing is performed to re-create tables in structured form before analysis, then data can be made available for processing, but significant manual work is required
Solution Approach 1:
The patent performs preliminary action by automatically creating structured table data before the analysis phase. The parser and extractor modules pre-process the static PDF to identify and structure all tabular data in advance, making it immediately available for subsequent analysis without requiring further manual intervention, thus reducing both complexity and manual work.
Solution Approach 2:
The system achieves universality by designing a multi-functional processing framework that can handle various document types and table structures through the same parser and extractor modules. This universal approach makes structured data available across different scenarios without requiring separate complex processing procedures for each case, reducing overall processing complexity.
Data Source
AI summary
Computerized systems and methods for identifying a table in a document include: removing from a document content other than text characters and associated size, position and format information; converting each text character into a block covering the corresponding text character; converting each page of the document into a corresponding image file; drawing a set of horizontal lines spanning a width of the document, each block super-scored and under-scored by at least one of the horizontal lines; drawing a set of vertical lines spanning all or a portion of a length of the document; removing a subset of redundant vertical lines; and determining, based on the set of horizontal lines and the subset of vertical lines, (i) a set of table coordinates corresponding to a table in the document, and (ii) one or more sets of cell coordinates corresponding to one or more cells in the table.