Deep Neural Network for Automatic Key-Value Extraction from Document Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional document-image data-capture systems require manual creation of templates for each document layout, making them time-consuming and expensive, especially when processing large volumes of documents with different layouts.
Innovation Solution
A system using deep neural networks to automatically create extraction templates with spatial coordinates and semantic labels, processing document images to extract key/value pairs without human intervention, by employing object detection and preprocessing techniques to enhance the recognition of textually encoded segments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual template creation is used for each document layout, then extraction accuracy is maintained, but processing time and cost increase significantly
Solution Approach 1:
The system enables automatic template creation where the deep neural network autonomously processes document images to generate extraction templates without human intervention. The network learns document layouts and creates templates automatically, eliminating the need for manual template creation while maintaining extraction accuracy across different document types.
Solution Approach 2:
The invention transforms the template creation process from a manual parameter-setting task to an automated learning process. By changing the parameters from manual coordinates and labels to neural network weights and biases, the system automatically adapts to different document layouts and creates appropriate templates dynamically.
2Measurement precision
If manual template creation is used for each document layout, then template accuracy is ensured, but system complexity increases
Solution Approach 1:
The invention replaces the mechanical manual process of template creation with an intelligent automated system. Instead of manually defining coordinates and labels, the deep neural network automatically learns document structures and generates templates, reducing system complexity despite the advanced technology involved.
Solution Approach 2:
The deep neural network serves multiple functions: it processes document images, detects layouts, creates templates, and extracts key/value pairs. This multi-functionality consolidates what would otherwise require separate manual operations into a single automated system, reducing overall system complexity.
3Productivity
If automatic template creation is implemented, then processing efficiency improves, but recognition accuracy for text segments may deteriorate
Solution Approach 1:
The system performs preliminary preprocessing of document images before template creation and extraction. This includes image enhancement, normalization, and feature extraction that prepare the input data for the neural network, ensuring high recognition accuracy while maintaining processing efficiency.
Solution Approach 2:
The deep neural network acts as an intermediary between the raw document image and the extraction process. It automatically detects text segments, determines their semantic meaning, and maps them to template fields, bridging the gap between image processing and accurate data extraction without requiring manual intervention.
Data Source
AI summary
Key/Value pairs, each comprising a keyword string and an associated value, are extracted automatically from a document image. Each document image has a plurality of pixels with each pixel having a plurality of bits. A first subset of the plurality of bits for each pixel represents information corresponding to the document image. The document image is processed to add information to a second subset of the plurality of bits for each pixel. The information added to the second subset alters the appearance of the document image in a manner that facilitates semantic recognition of textually encoded segments within the document image by a Deep Neural Network (DNN) trained to recognize images within image documents. The DNN detects groupings of text segments within detected spatial templates within the document image. The text segments are mapped to known string values to generate the keyword strings and associated values.