Semantic Data Column Classification Using ML Fit Scoring

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing methods for semantically classifying sets of data elements, such as using regular expressions or lookup tables, are resource-intensive and may not accurately represent the semantic type of the entire set of data elements.

Innovation Solution

A machine learning model is applied to predict semantic types for individual data elements within a data column, determining a semantic fit score for candidate types, and selecting the most appropriate type based on this score to improve data management operations like schema matching and data cleansing.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If regular expressions or lookup tables are used to semantically classify data elements, then classification can be performed, but resource consumption increases

Engineering Contradiction:

Improvesemantic classification accuracyVSAvoidresource consumption

Core Design Contradiction:

ReliabilityVSUse of energy by moving object

Solution Approach 1:

The patent replaces traditional mechanical classification methods (regular expressions and lookup tables) with a machine learning-based semantic classification system. The machine learning model learns semantic patterns from training data and automatically classifies data elements, substituting the rule-based mechanical approach with an intelligent system that reduces resource consumption while maintaining or improving classification accuracy.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Measurement precision

If regular expressions or lookup tables are used to determine semantic types of individual data elements, then classification can be performed, but the semantic type may not represent the set of data elements as a whole

Engineering Contradiction:

Improvesemantic type accuracyVSAvoidrepresentativeness of set semantic type

Core Design Contradiction:

Measurement precisionVSReliability

Solution Approach 1:

The patent merges the classification results of multiple individual data elements to determine the semantic type of the entire set. The machine learning model classifies each data element and then aggregates these classifications, using voting or frequency-based methods to determine the dominant semantic type that represents the set as a whole, ensuring both individual accuracy and set-level representativeness.

Inventive Principle:

Principle #5Merging (Combining)

Solution Approach 2:

The system incorporates feedback mechanisms where the classification results of individual data elements are aggregated and used to refine the determination of the set's semantic type. The model learns from the collective patterns across multiple elements, adjusting its classifications to better represent the overall semantic characteristics of the data set.

Inventive Principle:

Principle #23Feedback

Data Source

PatentUS12518106B2Semantically classifying sets of data elements

Publication Date: 2026.01.06 ORACLE INT CORP

AI summary

One or more embodiments include operations associated with semantic classification of data columns. The operations may include receiving a set of data elements corresponding to a data column to be semantically classified, applying a machine learning model to the set of data elements to predict a set of candidate semantic types for the set of data elements, selecting a particular semantic type from the set of candidate semantic types based at least in part on a semantic fit score corresponding to the particular semantic type predicted by the machine learning model, and presenting the particular semantic type as a recommended semantic classification for the data column.