Machine Learning Model for Document Item Value Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for extracting item values from documents with varying layouts are laborious and inefficient, as they require defining OCR read positions for each layout, making it difficult to automatically extract information from semi-fixed-format forms with different formats.
Innovation Solution
An information processing system that uses machine learning to generate a trained model based on positional relationships between candidate character strings and item keywords, allowing for the automatic extraction of item values from documents with unfixed layouts by determining the appropriateness of character strings as item values.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual definition of extraction rules is used for each layout, then extraction accuracy is improved, but processing time and labor requirements increase
Solution Approach 1:
The patent replaces manual mechanical definition of extraction rules with an automated machine learning system. The ML model automatically learns extraction patterns from training data and applies them to new documents, eliminating the need for manual rule creation while maintaining high extraction accuracy across varying layouts.
Solution Approach 2:
The patent changes the approach from fixed manual parameters (extraction rules defined by users) to adaptive learned parameters (model weights and features automatically adjusted during training). This allows the system to automatically adapt to different document layouts without manual reconfiguration.
2Reliability
If manual definition of extraction rules is used, then extraction reliability is improved, but device complexity and operational burden increase
Solution Approach 1:
The system performs self-service by automatically learning extraction patterns from training data without requiring manual rule definition. The machine learning model autonomously identifies relevant features and relationships, reducing both the complexity of manual configuration and the burden on users while maintaining reliable extraction.
3Productivity
If automatic extraction without layout definition is used, then processing speed is improved, but extraction accuracy deteriorates
Solution Approach 1:
The patent applies preliminary action by training the machine learning model on a comprehensive set of training data representing various document layouts before actual extraction occurs. This pre-training enables the model to accurately handle diverse layouts during production, achieving both high speed and high accuracy without manual per-layout configuration.
Data Source
AI summary
An information processing system includes circuitry that: acquires a character recognition result that is a result of character recognition performed on a target image; extracts, from the character recognition result of the target image, a plurality of candidate character strings that are candidates of an item value of an extraction target item; generates, for each of the plurality of candidate character strings, a feature quantity based on positional relationships between the candidate character string and a plurality of item keywords in the target image, the plurality of item keywords being keyword word strings for use in extraction of the item value of the extraction target item; stores a trained model in a memory; and inputs the feature quantity of each of the plurality of candidate character strings in the target image to the trained model so as to extract the item value of the extraction target item.