Automated Document Markup Using Historical Data Association
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies face challenges in efficiently extracting data from historical documents and training machine learning models due to the need for extensive manual markup and the discrepancies in data input, formatting, and errors.
Innovation Solution
The implementation describes a method and system for automatically identifying data in historical documents and training machine learning models by receiving a document, identifying corresponding records in a data structure, determining degrees of association between entries and referenced information, updating these associations, and selecting those that satisfy a criterion to generate an initial markup.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual markup is used to extract data from historical documents, then data extraction accuracy is improved, but time consumption and operational complexity increase significantly
Solution Approach 1:
The system performs preliminary actions by automatically generating initial markup from historical documents using machine learning models before human operators review them. This pre-processing step creates a draft markup that captures most extractable data, reducing the time operators need to spend on manual extraction while maintaining accuracy through subsequent review and correction of the pre-generated markup.
Solution Approach 2:
The system introduces an intermediary machine learning model that acts as a mediator between historical documents and final extracted data. The model generates initial markup that serves as an intermediate product, which is then refined by human operators. This intermediary step automates the time-consuming portions of data extraction while preserving human oversight for accuracy-critical decisions.
2Reliability
If extensive manual markup is performed on historical documents, then training data quality is improved, but operational complexity and resource requirements increase
Solution Approach 1:
The system enables self-service by allowing machine learning models to automatically generate markup for historical documents without requiring extensive manual intervention. The models learn from previously processed documents and autonomously create initial markup for new documents, reducing operational complexity while maintaining training data quality through iterative learning and automatic refinement processes.
Solution Approach 2:
The system applies partial action by having machine learning models generate only the portions of markup that can be automatically extracted with high confidence, leaving ambiguous or complex cases for human review. This approach produces sufficient training data quality by ensuring accurate markup for clear cases while reducing operational complexity by minimizing human involvement in straightforward extractions.
3Measurement precision
If machine learning models are trained using manually marked up documents, then model accuracy is improved, but training time and computational resources increase
Solution Approach 1:
The system performs preliminary action by pre-processing historical documents to generate initial markup using existing machine learning models before training new models. This pre-generated markup serves as ready-to-use training data, eliminating the need to create training data from scratch and significantly reducing training time while maintaining sufficient accuracy through the quality of pre-processed markup.
Solution Approach 2:
The system uses copying by replicating markup patterns and structures from previously processed documents to create training data for new models. Instead of manually creating unique training examples, the system copies and adapts proven markup patterns from historical documents, reducing training time and computational resources while preserving model accuracy through pattern consistency.
4Productivity
If automated markup is implemented without manual review, then productivity is improved, but data extraction accuracy may deteriorate
Solution Approach 1:
The system applies partial action by implementing automated markup only for portions of documents where machine learning models can extract data with high confidence, while routing ambiguous or complex cases to human reviewers. This hybrid approach maintains high productivity by automating straightforward extractions while preserving accuracy through selective human review of uncertain cases.
Solution Approach 2:
The system implements feedback loops where human reviewers correct and verify automated markup, and these corrections are fed back to retrain and improve the machine learning models. This continuous feedback process allows the system to maintain high productivity through automation while progressively improving data extraction accuracy over time as models learn from corrected examples.
Data Source
AI summary
A method of extracting information from documents includes: receiving a document; identifying, in a data structure, a record corresponding to the document, the record comprising one or more entries, each entry containing data referencing a respective item of information extracted from a specified location of the document; for each entry of the record, determining a degree of association between the entry and an item of information referenced by the entry; selecting, among a plurality of degrees of association between entries of the data structure and corresponding character strings, a set of degrees of association whose aggregate degree of association satisfies a criterion; and training, using the set of degrees of association, a machine learning model to extract information from new documents.