Document Key-Value Extraction With Rule-Based ML Post-Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies for extracting key-value pairs from documents such as invoices, purchase orders, packing slip, bills of lading, and contracts are labor-intensive, time-consuming, and prone to errors, especially when dealing with diverse document types and layouts.
Innovation Solution
A machine learning-based approach for key-value extraction that includes data preparation, feature engineering, and a two-step extraction process, utilizing OCR and XGBoost classification to accurately identify and classify key-value pairs in documents, with flexibility for adding new fields and languages.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual extraction methods are used, then accuracy can be maintained, but labor intensity and processing time increase significantly
Solution Approach 1:
The patent replaces manual mechanical extraction processes with an automated machine learning system. The system uses OCR (optical character recognition) to convert document images to text, followed by XGBoost classification models to automatically identify and extract key-value pairs, eliminating the need for manual data entry while maintaining high accuracy.
Solution Approach 2:
The system performs self-learning through training on labeled documents, where the machine learning model automatically improves its extraction capabilities. The model is trained on document images with annotated key-value pairs, enabling it to autonomously extract information from new documents without human intervention.
2Measurement precision
If manual extraction methods are used, then accuracy can be maintained, but time consumption increases
Solution Approach 1:
The patent replaces time-consuming manual extraction with automated computer vision and machine learning systems. The OCR engine rapidly converts document images to text, and the XGBoost classifier immediately processes the text to extract key-value pairs, reducing processing time from minutes to seconds while maintaining accuracy.
Solution Approach 2:
The system performs preliminary processing by pre-training the machine learning model on a comprehensive dataset of labeled documents. This pre-training enables the model to quickly and accurately extract information from new documents without requiring time-consuming manual analysis during the extraction process.
3Extent of automation
If existing extraction technologies are used, then processing can be automated, but error rates increase
Solution Approach 1:
The system incorporates feedback mechanisms where the extracted key-value pairs are validated against the original document images. The machine learning model continuously learns from corrections and feedback loops, improving its accuracy over time. The system can identify and correct extraction errors by comparing predicted values with actual document content.
Solution Approach 2:
The patent replaces error-prone rule-based extraction systems with neural network-based machine learning models. The XGBoost classifier and OCR engine work together to provide more reliable extraction, reducing errors caused by rigid rule-based approaches while maintaining high automation levels.
4Measurement precision
If the system is designed for specific document types, then accuracy improves, but adaptability to other document types decreases
Solution Approach 1:
The patent creates a universal machine learning system that can extract key-value pairs from multiple document types including invoices, purchase orders, packing slips, bills of lading, and contracts. The XGBoost model is trained on diverse document examples, enabling it to generalize and accurately extract information from new document types without requiring retraining or manual configuration for each document type.
Solution Approach 2:
The system adapts to different document types by dynamically adjusting extraction parameters and classification categories. The machine learning model can be fine-tuned by adding new document types to the training data, allowing the system to maintain high accuracy across diverse document formats while preserving versatility.
Data Source
AI summary
A method to improve, post-extraction, classification accuracy of key-values after a machine-learning model has been applied to documents, according to one embodiment, comprises receiving a collection of document images, creating an input data set from the collection, applying a classification model to the input data set that generates an initial set of entity predictions, and filtering the initial set of entity predictions that generates a revised set of entity predictions. The filtering the initial set of entity predictions further comprises applying at least a plurality of rules to the initial set of entity predictions. The plurality of rules comprises a first rule corresponding to treating each individual entity as unique, and a second rule corresponding to treating a single document as unique.