Unstructured Document Event Extraction via ML Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current systems are inefficient in automatically extracting and representing information from unstructured documents in a structured format, particularly for financial events within large volumes of unstructured textual data, such as corporate press releases, which requires manual effort and lacks supplemental information retrieval.
Innovation Solution
The system identifies and classifies events from unstructured text documents by generating document features using various schemas, applying trained classifiers to compute probability values, and representing the information in a structured format based on classification scores, along with assigning confidence scores for accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual parsing is used to identify events in unstructured documents, then information accuracy is maintained, but time consumption and labor effort increase significantly
Solution Approach 1:
The system enables automatic self-service extraction of events and attributes from unstructured documents using machine learning classifiers. The trained classifiers automatically identify and extract financial events, entities, and attributes without human intervention, resolving the contradiction by providing both speed (automatic processing) and accuracy (trained model predictions).
Solution Approach 2:
The patent replaces manual mechanical parsing with automated computational systems. Machine learning classifiers and probability estimation models substitute human analysts, automatically processing documents to extract events and attributes while maintaining accuracy through trained algorithms and confidence scoring.
2Productivity
If automated classification systems are implemented, then processing speed increases, but system complexity increases
Solution Approach 1:
The system segments the complex task of document analysis into distinct components: event detection, attribute extraction, classification, and confidence scoring. Each component is handled by specialized classifiers trained on specific aspects, making the overall system more manageable and maintainable while achieving high processing speed.
Solution Approach 2:
The patent creates a universal classification framework that handles multiple event types (dividends, earnings, mergers, etc.) and attributes through a single integrated system. The same architectural pattern and classification approach can be applied across different document types and event categories, reducing overall system complexity through reusability.
3Measurement precision
If comprehensive feature extraction is performed, then classification accuracy improves, but computational resources required increase
Solution Approach 1:
The system extracts only the most relevant features and attributes necessary for accurate classification rather than processing all possible document elements. The confidence scoring mechanism allows the system to identify high-confidence extractions first, reducing computational resources needed while maintaining accuracy for critical information.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Systems and techniques are disclosed for representing information included in unstructured text documents into a structured format. The systems and techniques identify events and information associated with the events in unstructured documents, classify the identified events and information, and represent the identified events and information in a structured format based on a computed classification score. The systems and techniques may also assign a confidence score to identified events, compare the confidence score associated with events to a confidence score associated with a trained confidence model, and represent the identified events and information associated with the events in a structured format based on the comparison.