Data Catalog Label Clustering for Misassigned Field Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing enterprise data catalogs lack a means to identify labels incorrectly assigned to data set fields, leading to potential misclassification and compliance risks due to improperly trained machine learning processes or misunderstood ontologies.
Innovation Solution
A computer-implemented method that represents data set fields using a set of characteristics, clusters them based on similarity, and identifies clusters with inconsistent label assignments, presenting suspect labels to users for review.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If machine learning processes are used to automatically assign labels to data set fields, then productivity is improved, but reliability deteriorates due to potential misclassification from improperly trained models or misunderstood ontologies
Solution Approach 1:
The system implements feedback by detecting inconsistencies in automatically assigned labels through clustering analysis. It identifies clusters of data set fields with the same label that exhibit high diversity in their data characteristics, then flags these suspect labels for review, creating a feedback loop that improves future labeling accuracy without reducing automation
Solution Approach 2:
The patent replaces manual label verification (mechanical human review) with an automated detection system using clustering algorithms and diversity metrics. This substitution maintains high productivity while improving reliability by using computational methods to identify labeling errors that automated systems cannot detect through traditional validation
2Reliability
If comprehensive manual review of all labeled data set fields is performed, then reliability is improved, but productivity deteriorates due to time-consuming analysis
Solution Approach 1:
Instead of reviewing all labels comprehensively, the system performs partial action by selectively identifying only those labels that exhibit inconsistencies through clustering analysis. It flags suspect labels for review while leaving confident labels untouched, achieving high reliability verification without the productivity loss of complete manual review
Solution Approach 2:
The system enables self-service by allowing the data catalog to automatically identify and flag its own labeling errors through internal consistency checks. The clustering-based detection mechanism allows the system to service itself by generating quality assurance reports without requiring external manual auditing of all data
3Measurement precision
If clustering algorithms are used to identify suspect labels, then measurement precision is improved, but device complexity increases due to additional computational requirements
Solution Approach 1:
The system changes parameters by using diversity metrics within clusters as the key indicator for detecting suspect labels. Instead of complex classification models, it transforms the problem into measuring data diversity within groups of similarly-labeled fields, achieving high measurement precision through simpler computational parameters that can be calculated efficiently
Data Source
AI summary
Described are techniques for detecting labels incorrectly assigned to data set fields. The data of each data set field, such as those data set fields assigned to the same label, are represented using a set of characteristics. The data set fields are then clustered into clusters based on the characteristics of the data of the data set fields. Those clusters of data set fields with a homogeneity (being assigned the same label) that exceeds a first threshold value and is below a second threshold value are identified. One or labels assigned to the data set fields of the identified clusters are identified as being suspect for incorrect assignments by having a frequency below a third threshold value (e.g., 3%), which may be user-designated. The label(s) identified as being suspect for incorrect assignment are then presented to a user for review.