Document Correction History for Accurate OCR on Ambiguous Formats
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing character string extraction techniques struggle with accurately extracting strings from documents with company-specific content or ambiguous formats, such as invoices lacking a written year, leading to decreased recognition accuracy.
Innovation Solution
An information processing apparatus and method that uses circuitry to receive correction content, store document correction history, and match input documents with identical formats to correct character strings based on stored correction content when a threshold is met, utilizing document type, company information, and statement table layouts for accurate extraction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If AI training is performed on a large number of documents to gain high extraction accuracy, then character string extraction accuracy is improved, but the system becomes unable to handle company-specific content and ambiguous formats effectively
Solution Approach 1:
The patent segments the document processing into distinct categories (structured documents with fixed formats, semi-structured documents with similar formats, and unstructured documents with diverse formats) and applies different processing strategies to each segment. This allows the system to maintain high accuracy for each document type while avoiding the limitation of trying to handle all formats uniformly with a single AI model.
Solution Approach 2:
The patent implements dynamic processing by detecting the document format type and automatically selecting the appropriate extraction method. The system transitions from static AI training on fixed formats to dynamic adaptation that can handle multiple format types through format detection and selective application of structured/semi-structured/unstructured OCR techniques.
2Productivity
If traditional OCR methods are used for diverse document formats, then processing speed is maintained, but extraction accuracy decreases for company-specific content and ambiguous formats
Solution Approach 1:
The system dynamically selects between traditional OCR methods and AI-based extraction methods based on the detected document format. For structured and semi-structured documents with recognizable formats, traditional OCR maintains processing speed. For unstructured documents with ambiguous formats, AI-based extraction is activated to improve accuracy, thus optimizing the balance between speed and precision across different document types.
Solution Approach 2:
The patent changes the processing parameters and methods based on the document format type. Different extraction algorithms and parameter settings are applied to structured, semi-structured, and unstructured documents respectively, allowing the system to maintain high processing speed for familiar formats while achieving high accuracy for complex formats through parameter adaptation.
3Device complexity
If the system processes all document types with the same extraction method, then device complexity is reduced, but extraction accuracy decreases for specific document formats
Solution Approach 1:
The patent segments the extraction process into distinct methods for structured, semi-structured, and unstructured documents. By detecting the document type and applying the appropriate specialized method, the system achieves high accuracy for each format type while maintaining a relatively simple overall architecture based on format detection and selective method application.
Data Source
AI summary
An information processing apparatus includes circuitry that: receives correction content indicating a change from a first character string extracted from first document data to a second character string; stores in a memory a document correction history representing the correction content of the first document data, and identical document information used for determining whether an input document has an identical format with the first document data; acquires second document data as the input document; extracts a third character string from the second document data; calculates a degree of match between the second document data and the identical document information; and when the degree of match is equal to or greater than a threshold value, and a comparison result between the third character string and the document correction history meets a predetermined condition, corrects the third character string based on the correction content represented by the document correction history.