Neural Network Framework for Document Hierarchy Extraction

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional methods for extracting hierarchical relationships from documents are suboptimal, particularly in handling multi-level relationships and are resource-intensive, failing to accurately capture nested tree structures.

Innovation Solution

A method and system that utilize a neural network framework, trained with a multimodal approach combining textual and visual modalities, to classify and identify key entity fields and their corresponding values, forming a training dataset to accurately extract multi-level hierarchical relationships.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If conventional language processing models are used for information extraction, then the extraction process can be performed, but the models are unable to extract multi-level hierarchical relationships between entity fields and cause overburdening on computing resources

Engineering Contradiction:
Improveextraction accuracyVSAvoidcomputing resource burden
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent segments the hierarchical relationship extraction into multiple processing levels. The neural network framework processes data element relationships in a structured manner, identifying parent-child relationships at different hierarchical levels separately, which reduces computational complexity while maintaining extraction accuracy for multi-level hierarchies

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent introduces a trained neural network framework as an intermediary between the input document and the extracted hierarchical relationships. This intermediary has been pre-trained on hierarchical data structures, enabling it to efficiently identify relationships without requiring excessive computing resources during actual extraction operations

Inventive Principle:
Principle #24Intermediary (Mediator)

2Measurement precision

If conventional language processing models are used for information extraction, then the extraction process can be performed, but the models are time consuming

Engineering Contradiction:
Improveextraction accuracyVSAvoidextraction time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent applies preliminary action by pre-training the neural network framework on hierarchical data structures before actual extraction. The training phase establishes the model's understanding of hierarchical relationships in advance, so that during actual document processing, the extraction occurs rapidly without requiring extensive computational time

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The extraction process is segmented into efficient processing stages within the neural network framework, allowing rapid identification of hierarchical relationships without requiring time-consuming sequential analysis of all possible element combinations

Inventive Principle:
Principle #1Segmentation

Data Source

PatentUS20240320483A1Method and system for identifying hierarchical relationships between data elements of document
Publication Date: 2024.09.26 QUANTIPHI INC
  • US20240320483A1 patent drawing
  • US20240320483A1 patent drawing

AI summary

Disclosed is a method for identifying multi-level hierarchical relationships between data elements of a document, the method comprising receiving a plurality of sample documents each having a plurality of data elements arranged in a multi-level hierarchical data structure; classifying each of the plurality of data elements into a key entity field or a key field value based on a hierarchical relationship therebetween; identifying key entity fields, from among the classified key entity field of the plurality of data elements, having the hierarchical relationship therebetween; pairing the key entity field, with a corresponding key field value or an identified key entity field, to form a training dataset; and employing the training dataset on a neural network framework, having at least one of a textual modality or a visual modality, to identify the multi-level hierarchical relationships between the data elements of the document.