Neural Network Segmentation for Document Data Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems face challenges in efficiently processing electronic documents, such as invoices, due to the need for massive neural networks that require significant computing resources and long training times, and struggle to understand the semantics and relationships within these documents.
Innovation Solution
The use of an integrated neural network structure comprising fully connected neural networks that classify texts into canonical categories, leveraging relative spatial positions and hierarchical integration to reduce noise and improve classification accuracy, allowing for faster training and reduced computing resources.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a single massive neural network is used to extract information from electronic documents, then information extraction capability is improved, but computing resources required and training time increase significantly
Solution Approach 1:
The patent divides the single massive neural network into multiple smaller neural networks, each responsible for extracting specific types of information from electronic documents. This segmentation reduces the computational burden on each individual network while maintaining overall extraction capability, directly addressing the contradiction between extraction capability and computing resource consumption.
2Measurement precision
If a single massive neural network is used to extract information from electronic documents, then information extraction capability is improved, but training time increases significantly
Solution Approach 1:
By segmenting the neural network into multiple smaller specialized networks, each network can be trained independently and more efficiently on specific document features. This reduces the overall training time compared to training one massive network, while still achieving comprehensive information extraction capability.
3Device complexity
If a single neural network is used to process electronic documents, then processing is simplified, but understanding of text semantics and relationships deteriorates
Solution Approach 1:
The patent assigns different semantic understanding capabilities to different neural networks based on their specific functions. Each network is optimized to understand particular types of text relationships and semantics relevant to its extraction task, improving overall semantic understanding while maintaining processing efficiency through specialization.
Data Source
AI summary
Described herein is a computer-implemented method for automatic extraction of canonical data from an electronic document. The method comprises classifying a first text rectangle in an electronic document as a label and a second text rectangle as a value using a first machine learning algorithm. A first probability score of a likelihood of the first text rectangle corresponding to a first canonical category is determined using a second machine learning algorithm. A second probability score of a likelihood of the second text rectangle corresponding to a first canonical category is determined using a third machine learning algorithm. A relative spatial position of the second text rectangle relative to the first text rectangle is calculated. Based on the relative spatial position, the first probability score, and the second probability score, the first text rectangle, and the second text rectangle are classified into the first canonical category.