Document Type Identification Using Positional Relationship Features
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing document classification systems face challenges in accurately identifying document types with varying layouts, as they rely on fixed rules or features like word presence and position, which are not effective for documents with unfixed formats, such as invoices or receipts.
Innovation Solution
The system generates a trained model through machine learning using positional relationship feature quantities, including position, distance, size, and row features of frequently occurring words, to identify document types, enabling accurate classification even for documents with different layouts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If fixed rules or features like word presence and position are used for document classification, then the system is simple to implement, but it cannot accurately identify document types with varying layouts
Solution Approach 1:
The patent transforms the document classification approach by changing from fixed rule-based parameters to machine learning models that process multiple feature parameters including positional relationships, word frequencies, and layout structures. This allows the system to adapt to varying document layouts while maintaining identification accuracy.
Solution Approach 2:
The system employs dynamic machine learning models that can adapt to different document types and layouts rather than relying on static fixed rules. The models are trained on diverse document data and can dynamically adjust classification decisions based on the specific characteristics of each document.
2Adaptability or versatility
If machine learning models are used to handle documents with varying layouts, then document type identification accuracy improves, but the system complexity increases
Solution Approach 1:
The patent develops universal machine learning models that can handle multiple document types and layouts through a single unified system. The models are designed to process various document formats (invoices, receipts, contracts, etc.) using common feature extraction and classification mechanisms, reducing the need for separate specialized systems for each document type.
Solution Approach 2:
The system performs preliminary training of machine learning models using large datasets of annotated documents before deployment. This preliminary action prepares the models to handle diverse document layouts effectively, reducing the complexity of real-time classification decisions during actual operation.
3Measurement precision
If positional relationship feature quantities are extracted between frequently occurring words, then identification accuracy for unfixed layout documents improves, but the processing complexity increases
Solution Approach 1:
The patent extracts specific positional relationship features between frequently occurring words as key discriminative elements for document classification. By focusing on the relative positions of semantically important words rather than all text elements, the system reduces the dimensionality of feature space while maintaining or improving classification accuracy.
Solution Approach 2:
The system extracts positional relationships for only the most frequently occurring and semantically significant words rather than all words in the document. This partial action approach reduces processing complexity while capturing the essential structural characteristics needed for accurate document type identification.
Data Source
AI summary
An information processing system includes: circuitry that: acquires a character recognition result of an identification target image; stores a frequently occurring word string of a predetermined document type; detects the frequently occurring word string from the character recognition result of the identification target image to acquire information on a position of the frequently occurring word string in the identification target document; generates a feature quantity of the identification target document using the information on the position, the feature quantity including a positional relationship feature quantity related to a positional relationship between the frequently occurring word string and another word string in the identification target document; stores a trained model that identifies the predetermined document type; and inputs the feature quantity of the identification target document to the trained model to identify whether the identification target document is a document of the predetermined document type.