Neural Network Document Field Detection

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Conventional methods for detecting fields in unstructured electronic documents are inefficient, requiring numerous manually configured heuristics and templates, which are labor-intensive and lack flexibility, especially when dealing with varied document formats and confidential information.

Innovation Solution

The use of neural networks to analyze document layouts, classify documents into types, and detect fields using cluster-specific neural networks trained on a limited number of marked-up documents, allowing for client-side training and flexible field detection tailored to specific client needs.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of operation

If conventional heuristic-based methods are used for field detection, then manual configurability is achieved, but labor intensity and time consumption increase significantly

Engineering Contradiction:

Improvemanual configurabilityVSAvoidtime consumption

Core Design Contradiction:

Ease of operationVSLoss of time

Solution Approach 1:

The patent replaces manual heuristic configuration (mechanical system) with an automated neural network-based field detection system. The neural network automatically learns document structures and detects fields without requiring manual setup of heuristics, thereby eliminating the time-consuming manual configuration process while maintaining detection capability.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The neural network system performs self-service by automatically adapting to different document types and formats without human intervention. It learns from training data and autonomously detects fields in new documents, eliminating the need for continuous manual reconfiguration and reducing ongoing time investment.

Inventive Principle:

Principle #25Self-service

2Adaptability or versatility

If multiple manually configured templates are used to handle varied document formats, then document type coverage is improved, but system complexity and maintenance effort increase

Engineering Contradiction:

Improvedocument type coverageVSAvoidsystem complexity

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent implements a universal neural network-based field detection system that can handle multiple document types and formats through a single unified approach. Instead of maintaining separate templates for each document type, the neural network learns general document structures and adapts to specific formats automatically, providing multi-functionality without increasing system complexity.

Inventive Principle:

Principle #6Universality (Multi-functionality)

Solution Approach 2:

The system uses parameter changes in the neural network weights and configurations to adapt to different document types. Rather than changing the fundamental system structure or adding complex templates, the neural network adjusts its internal parameters during training and inference to accommodate varied document formats, maintaining simplicity while achieving versatility.

Inventive Principle:

Principle #35Parameter changes

3Measurement precision

If comprehensive manual markup is performed for training, then detection accuracy is improved, but exposure to confidential information and manual effort increase

Engineering Contradiction:

Improvedetection accuracyVSAvoidexposure to confidential information

Core Design Contradiction:

Measurement precisionVSObject-affected harmful factors

Solution Approach 1:

The neural network performs self-service learning by automatically extracting training signals from minimally marked-up documents. Instead of requiring comprehensive manual markup that exposes confidential information, the system uses limited annotations and autonomously learns document structures, reducing both manual effort and exposure to sensitive data while maintaining detection accuracy.

Inventive Principle:

Principle #25Self-service

4Device complexity

If a single neural network is used for all document types, then system simplicity is maintained, but detection precision for specific document formats decreases

Engineering Contradiction:

Improvesystem simplicityVSAvoiddetection precision

Core Design Contradiction:

Device complexityVSMeasurement precision

Solution Approach 1:

The patent segments the field detection task by implementing separate neural networks for different document types or formats. Each specialized network is trained on specific document types, allowing it to achieve high detection precision for that particular format. This segmentation maintains relative system simplicity through modular architecture while significantly improving detection precision compared to a single general-purpose network.

Inventive Principle:

Principle #1Segmentation

Data Source

PatentUS11816165B2Identification of fields in documents with neural networks without templates

Publication Date: 2023.11.14 ABBYY DEVELOPMENT INC

AI summary

Aspects of the disclosure provide for mechanisms for identification of fields in documents using neural networks. A method of the disclosure includes obtaining a layout of a document, the document having a plurality of fields, identifying the document, based on the layout, as belonging to a first type of documents of a plurality of identified types of documents, identifying a plurality of symbol sequences of the document, and processing, by a processing device, the plurality of symbol sequences of the document using a first neural network associated with the first type of documents to determine an association of a first field of the plurality of fields with a first symbol sequence of the plurality of symbol sequences of the document.