Document Layout Information Extraction via Image Conversion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional systems struggle to accurately extract data from files, especially those with varied formats and layouts, and are unable to replicate human-like reading and data extraction processes at scale.
Innovation Solution
The method involves converting a file into an image and using multiple models to extract information and generate outputs with coordinates, which are then merged to create a searchable output file organized by regions of interest.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional systems are used to extract data from files, then the extraction process is simple, but the accuracy of data extraction is poor and cannot handle varied formats and layouts
Solution Approach 1:
The patent segments the data extraction process into multiple specialized models: an OCR model for text recognition, a layout analysis model for structure identification, and an information extraction model for data retrieval. Each model handles a specific aspect of the extraction process, improving overall accuracy while managing complexity through functional decomposition
Solution Approach 2:
The patent introduces an intermediary step of converting files to images and generating bounding box coordinates before final information extraction. This intermediary representation allows the system to accurately capture the spatial layout and structure of the document, enabling more precise data extraction from varied formats
2Measurement precision
If multiple models are used to extract information with coordinates, then the extraction accuracy improves, but the processing time and computational resources increase
Solution Approach 1:
The patent performs preliminary actions by first converting the file to an image format and generating bounding box coordinates for all elements before the actual information extraction. This preliminary structuring of data with spatial information enables the subsequent extraction models to work more efficiently and accurately, reducing overall processing time despite the additional initial steps
3Adaptability or versatility
If files are converted to images and processed through multiple models, then the ability to extract structured information improves, but the complexity of the processing pipeline increases
Solution Approach 1:
The patent implements a universal processing pipeline that handles multiple file formats (PDF, images, documents) through a common architecture. The file conversion to image format serves as a universal intermediate representation that enables consistent processing across different input types, while the coordinated models provide multi-functional capabilities for various extraction needs
4Ease of operation
If coordinates are generated for all elements and regions of interest, then the organization and searchability of extracted information improves, but the computational overhead increases
Solution Approach 1:
The patent extracts only the necessary coordinate information for regions of interest rather than processing all elements uniformly. By identifying and extracting only the relevant bounding boxes and spatial information needed for the specific extraction task, the system improves organization and searchability while reducing unnecessary computational overhead from processing irrelevant elements
Data Source
AI summary
A method includes receiving a file in a first format and converting the file into an image. The file includes information and a plurality of regions of interest (ROIs). The method also includes generating a first output that includes a first set of information and a first set of coordinates of the first set of information in the image. The method also includes generating a second output including a second set of coordinates for each ROI of the plurality of ROIs and generating an output file in a second format. The output file includes a plurality of sections each corresponding to an ROI and included in the output file based on coordinates of an ROI in the second set of coordinates. The method also includes populating each section of the plurality of sections in the output file with a portion of the information determined to correspond with the respective section.