Metadata Extraction Using Cone Orientation and Area Relevance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for extracting metadata from semi-structured documents, such as invoices and HTML documents, face challenges due to inconsistent structural elements and layouts, leading to poor performance in identifying relevant metadata labels and values, especially with OCR errors and varying document formats.
Innovation Solution
A method utilizing area and cone orientation as relevance measures between words/phrases, combined with a training mechanism to correct OCR errors and handle user corrections in both online and offline modes, involves converting documents to a readable format, marking potential labels and values, and mapping relevant pairs using adjustable parameters to achieve accurate metadata extraction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If distance-based relevance measure is used to identify metadata labels and values, then the method is simple to implement, but it performs poorly and is not always correct in semi-structured documents
Solution Approach 1:
The patent changes the relevance measurement parameter from simple distance to a composite parameter considering area, cone orientation, and spatial relationships. This allows the system to accurately identify label-value pairs in semi-structured documents with varying layouts while maintaining computational feasibility through defined algorithms for area calculation and cone orientation measurement.
2Measurement precision
If boundaries are defined for each value to control mappings, then false positive mappings are reduced, but the approach cannot be generalized across different layouts
Solution Approach 1:
The patent creates a universal relevance measurement system based on area and cone orientation that works across multiple document layouts and formats. The method defines generalizable spatial relationships and angular measurements that can identify label-value pairs in invoices, bills, and other semi-structured documents regardless of specific layout variations, eliminating the need for layout-specific boundary definitions.
3Ease of operation
If OCR is used to convert scanned documents to readable format, then document accessibility is improved, but OCR errors affect coordinate interpretation accuracy
Solution Approach 1:
The patent implements a feedback mechanism where the extracted metadata and spatial relationships are used to verify and correct OCR errors. The system cross-validates coordinate interpretations against the structural patterns identified through area and cone orientation analysis, allowing it to detect and correct OCR mistakes in coordinate data while maintaining the benefits of OCR-based text recognition.
Data Source
AI summary
A method of extracting meta-data from semi structured documents, by using area and cone orientation as relevance between words/phrases is described. It also a computer implemented system to handle OCR errors with respect to the coordinates interpreted for each word and user corrections both in online and offline mode. The method is carried out by the steps as follows: converting scanned or digital document to a readable format with coordinates using OCR; scanning the coordinates obtained through OCR for each character; marking all potential labels and values with a bounding box; searching for relevant labels for the particular value by using default control parameters and adjusting trainable parameters; mapping a cone region for the labels and values using the upper and lower angles along x-axis and the scope box and formulating the score area to get the confidence percentage which is used as measure to extract all relevant label-value pairs.