Context-Aware Text Extraction via Graph-Based Entity Refinement

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current text extraction techniques in OCR systems lack context awareness and historical dependency consideration, leading to suboptimal extraction of structured documents like financial forms and receipts, where entities are interdependent.

Innovation Solution

The system employs a context-aware approach by constructing a graph using nodes and edges from OCR outputs, adjusting confidence values based on historical user data, and calculating cross-entity likelihoods to determine the best clique for accurate entity extraction.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If independent entity extraction is used, then extraction speed is maintained, but extraction accuracy deteriorates due to lack of context awareness

Engineering Contradiction:

Improveextraction accuracyVSAvoidsystem complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The system segments the extraction process into two distinct phases: (1) independent entity extraction using OCR and machine learning classifiers, and (2) context-aware refinement using graph analysis. This segmentation allows the system to maintain speed benefits of independent extraction while adding context-awareness in a separate refinement stage, resolving the contradiction between accuracy and complexity.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The system performs preliminary independent entity extraction before conducting context-aware refinement. By extracting entities independently first and then refining them using graph-based context analysis, the system prepares data in advance for the more complex context-aware processing, improving accuracy without completely redesigning the extraction pipeline.

Inventive Principle:

Principle #10Preliminary action

2Measurement precision

If context-aware processing is added, then extraction accuracy improves, but processing time increases

Engineering Contradiction:

Improveextraction accuracyVSAvoidprocessing time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The system divides processing into fast independent extraction and slower context-aware refinement stages, allowing time-critical operations to proceed quickly while applying computationally intensive context analysis only where needed to improve accuracy.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The system applies context-aware graph analysis selectively rather than to all entities uniformly. By focusing computational resources on entities where context information is most valuable and using heuristics to identify promising candidates for refinement, the system achieves accuracy improvements without proportionally increasing processing time for all entities.

Inventive Principle:

Principle #16Partial or excessive action

3Reliability

If historical data integration is implemented, then interdependency modeling improves, but data processing complexity increases

Engineering Contradiction:

Improveinterdependency modelingVSAvoiddata processing complexity

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The system introduces a graph structure as an intermediary representation that captures entity interdependencies and historical relationships. This graph serves as a mediator between raw historical data and entity extraction results, organizing complex data relationships in a structured format that simplifies processing and querying while improving reliability of interdependency modeling.

Inventive Principle:

Principle #24Intermediary (Mediator)

Data Source

PatentUS11651605B1Systems and methods for context-aware text extraction

Publication Date: 2023.05.16 INTUIT INC

AI summary

Systems and methods are provided to perform context-aware text extraction.