Zero-Shot Audit Information Labelling for Multi-Risk Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current audit processes rely on manual and subjective methods for assigning risk labels, which are time-consuming and prone to human error, often failing to capture the full scope of risks described in audit issues.
Innovation Solution
A zero-shot intelligent classifier (ZINC) is used to automatically label audit issues by combining issue descriptions with risk taxonomy hypotheses, applying a generative large-language model to determine relevant sub-risks, and providing multi-label classification without requiring training data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual review and assignment of risk labels is used, then auditor expertise and judgment are applied, but the process is highly time-consuming and subjective
Solution Approach 1:
The patent replaces the manual mechanical process of auditor review and label assignment with an automated natural language processing system. The system uses machine learning models to analyze issue descriptions and automatically assign risk labels, eliminating the time-consuming manual mechanical review process while maintaining classification accuracy through algorithmic pattern recognition.
Solution Approach 2:
The system enables self-service automated risk classification where the audit issues are automatically processed and labeled without requiring auditor intervention for each individual issue. The NLP system independently performs text analysis, risk assessment, and label assignment, allowing the system to serve itself in the classification task.
2Adaptability or versatility
If manual risk label assignment is used, then a single risk label can be assigned, but the full scope of risks is not captured
Solution Approach 1:
The patent segments the risk classification process into multiple independent analysis components that evaluate different aspects of the issue description. The system generates multiple risk labels by analyzing the text through different NLP models and approaches, then aggregates these results to provide a comprehensive multi-label classification that captures the full scope of risks.
Solution Approach 2:
The NLP system performs multiple functions simultaneously: it analyzes text semantics, identifies risk patterns, determines label relevance, and generates multiple risk classifications. This multi-functional approach allows a single system to handle various risk types and classification requirements, providing versatile and comprehensive risk labeling.
3Measurement precision
If auditor familiarity with entire risk taxonomy is required, then expert judgment is utilized, but human error increases
Solution Approach 1:
The patent replaces the human auditor's mechanical cognitive process of reviewing and understanding the entire risk taxonomy with an automated NLP system. The system has inherent knowledge of the risk taxonomy structure and uses algorithmic processing to match issue descriptions with appropriate risk labels, eliminating human error while maintaining high accuracy through consistent application of classification rules.
Data Source
AI summary
A zero-shot classifier can be used for the automatic labelling of audit information. An issue description in the audit information is compared to each of a plurality of risk/sub-risk descriptions using a zero-shot classifier in order to determine a plurality of risk/sub-risks that are most relevant to the issue description.