Neural Network Data Extraction from Electronic Documents

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current software tools fail to accurately extract data from electronic documents, requiring manual transcription and significant labor, which delays information ingestion and is not applicable to various textual information retrieval applications.

Innovation Solution

A computerized method using neural networks to extract data values from semi-structured electronic documents by determining pixel coordinates and extracting field identifiers and data values, with reinforcement learning feedback, and storing the data in a database for display on user devices.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If manual transcription is used to extract data from electronic documents, then accuracy can be maintained, but labor hours and time consumption increase significantly

Engineering Contradiction:

Improvedata extraction accuracyVSAvoidinformation ingestion speed

Core Design Contradiction:

Measurement precisionVSProductivity

Solution Approach 1:

The patent replaces the mechanical manual transcription process with an automated neural network-based optical character recognition (OCR) system. The neural network analyzes scanned document images, identifies text regions, and extracts data fields automatically, eliminating the need for manual typing while maintaining high accuracy through intelligent pattern recognition and learning algorithms.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The patent introduces a neural network as an intermediary between the scanned document image and the extracted data. This intermediary layer processes the image data, recognizes text patterns, and outputs structured information, serving as a bridge that automates the transcription process while preserving accuracy through sophisticated pattern matching and interpretation.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Ease of operation

If general software tools are used for data extraction, then ease of operation is maintained, but extraction accuracy fails to meet requirements

Engineering Contradiction:

Improvesoftware usabilityVSAvoiddata extraction accuracy

Core Design Contradiction:

Ease of operationVSMeasurement precision

Solution Approach 1:

The patent changes the fundamental parameters of the data extraction system by implementing a neural network with adjustable weights and biases that learn from training data. This allows the system to adapt to different document formats, layouts, and content types, maintaining ease of operation while achieving high extraction accuracy through continuous learning and parameter optimization.

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The patent transforms the static, rule-based software tools into a dynamic system that adapts to different document types and extraction requirements. The neural network continuously learns from new data, adjusts its internal parameters, and improves its performance over time, enabling both ease of operation and high accuracy across diverse applications.

Inventive Principle:

Principle #15Dynamics

3Adaptability or versatility

If manual data extraction is performed, then adaptability to various document types can be handled, but significant labor hours are required

Engineering Contradiction:

Improveapplicability to various textual information retrieval applicationsVSAvoidlabor hours for transcription

Core Design Contradiction:

Adaptability or versatilityVSLoss of time

Solution Approach 1:

The patent replaces manual data extraction operations with an automated neural network system that can process various document types without human intervention. The system handles different layouts, formats, and content structures by learning from diverse training data, eliminating the need for manual transcription while maintaining versatility across multiple applications.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The patent creates a universal data extraction system based on neural networks that can handle multiple document types and extraction tasks through a single platform. The system's ability to learn from various training datasets enables it to adapt to different applications, from invoice processing to form data extraction, without requiring separate manual efforts for each document type.

Inventive Principle:

Principle #6Universality (Multi-functionality)

Data Source

PatentUS11790215B2Systems and methods for object detection and data extraction using neural networks

Publication Date: 2023.10.17 FMR CORP

AI summary

Systems and methods for extracting data values from electronic documents using neural networks. The method includes receiving an electronic document having data values and associated field identifiers, determining pixel coordinates corresponding to the field identifiers using a first neural network, and extracting the field identifiers located at the pixel coordinates using a second neural network. The method also includes, for each of the field identifiers, calculating pixel coordinates on the electronic document corresponding to a data value associated with the field identifier using a third neural network and extracting the data value located at the calculated pixel coordinates using the second neural network. The method further includes, for each of the data values, generating a record in a data structure, the record including the extracted value and the extracted field identifier. The method also includes storing the data structure including the records in a database.