Convolutional Neural Network Structured Document Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for processing structured documents lack effective feature extraction and classification due to the inability to accurately leverage relative content positions in rendered forms, leading to suboptimal classification results.
Innovation Solution
A method involving the generation of a matrix representation of structured documents using numeric embeddings, processed through a convolutional neural network, which includes convolutional and classification subnetworks to extract features and generate accurate classification data based on relative content positions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional methods are used to process structured documents, then the processing approach is simple, but the classification accuracy is insufficient due to inability to leverage relative content positions
Solution Approach 1:
The patent replaces traditional mechanical document processing methods with a neural network-based system. Specifically, it uses a convolutional neural network (CNN) to process structured documents, substituting conventional text analysis algorithms with a learned model that can automatically capture spatial relationships and content positions, thereby improving classification accuracy while managing complexity through automated feature extraction.
Solution Approach 2:
The patent transforms the document processing problem by introducing a spatial dimension through the use of a grid-based matrix representation. By mapping document content to a two-dimensional grid structure where cells represent spatial positions, the system enables the CNN to leverage relative content positions and spatial relationships, adding a dimensional aspect that traditional linear processing methods lack.
2Measurement precision
If feature extraction capabilities are enhanced to reflect relative content positions, then classification accuracy improves, but the processing complexity increases
Solution Approach 1:
The patent segments the document processing task into distinct components: a grid generation module that creates a matrix representation of the document layout, and a convolutional neural network module that processes this grid structure. This segmentation allows the system to separately handle spatial relationship extraction and classification, making the complex feature extraction process more manageable and interpretable while maintaining high classification accuracy.
Data Source
AI summary
Structured documents are processed using convolutional neural networks. For example, the processing can include receiving a rendered form of a structured document; mapping a grid of cells to the rendered form; assigning a respective numeric embedding to each cell in the grid, comprising, for each cell: identifying content in the structured document that corresponds to a portion of the rendered form that is mapped to the cell, mapping the identified content to a numeric embedding for the identified content, and assigning the numeric embedding for the identified content to the cell; generating a matrix representation of the structured document from the numeric embeddings assigned to the cells of the grids; and generating neural network features of the structured document by processing the matrix representation of the structured document through a subnetwork comprising one or more convolutional neural network layers.