Machine Learning Semantic Equivalence Detection in Query Plans

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing solutions for detecting equivalence in large-scale analytics engines are inadequate, particularly in identifying semantic equivalence between subexpressions, leading to computational redundancy and inefficiencies due to the complexity and scalability issues of current methods.

Innovation Solution

A general equivalence optimizer (GEqO) engine that applies a series of filters, including a schema filter, vector matching filter, equivalence model filter, and automated verifier, to efficiently identify semantically equivalent subexpressions, reducing computational workload by eliminating duplicative queries.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If exhaustive verification methods are used to detect semantic equivalence, then detection precision is improved, but computational time and complexity increase significantly

Engineering Contradiction:
Improvedetection precisionVSAvoidcomputational time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The verification process is segmented into multiple filtering stages: schema filter, vector matching filter, and equivalence model filter. Each stage processes a portion of the equivalence detection task, progressively narrowing down candidate subexpressions before applying the most computationally intensive verification methods only to the most promising candidates.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

Instead of applying full verification to all subexpression pairs, the system applies partial verification through progressive filtering. The schema filter and vector matching filter perform lighter checks on all candidates, while the equivalence model filter applies more intensive semantic equivalence checking only to pairs that pass earlier filters, reducing overall computational burden.

Inventive Principle:
Principle #16Partial or excessive action

2Productivity

If semantic equivalence detection is implemented to eliminate duplicative computations, then productivity is improved, but device complexity increases

Engineering Contradiction:
Improvecomputational efficiencyVSAvoidsystem complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The equivalence detection system is designed as a universal framework that can detect both syntactic and semantic equivalence across different types of subexpressions and query plans. The multi-stage filter architecture serves multiple functions: initial filtering, candidate generation, and verification, making the system adaptable to various equivalence detection scenarios without requiring separate specialized systems.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Solution Approach 2:

The verification system is structured with nested filtering stages where simpler filters are embedded within progressively more complex verification layers. The schema filter operates at the outermost level, followed by vector matching filter, and finally the equivalence model filter for semantic equivalence. This nested structure allows the system to manage complexity by organizing verification tasks in hierarchical layers.

Inventive Principle:
Principle #7Nested doll (Nesting)

Data Source

PatentEP4625199A1Machine learning accelerated semantic equivalence detection
Publication Date: 2025.10.01 MICROSOFT TECHNOLOGY LICENSING LLC
  • EP4625199A1 patent drawingFigure 1
  • EP4625199A1 patent drawingFigure 2
  • EP4625199A1 patent drawingFigure 3A

AI summary

Examples detect equivalent subexpressions within a computational workload. Examples include converting a query plan tree associated with a first subexpression into a matrix. The first subexpression is a portion of a database query from the computational workload. Each node in the query plan tree is represented as a row of the matrix. The matrix is converted into a first vector. The first subexpression is determined to be equivalent to a second subexpression by comparing the first vector to a second vector associated with the second subexpression. The comparison includes computing a distance between the first and second vectors that is lower than a distance threshold. The computational workload is modified, based on the determining, to perform the first subexpression and exclude performance of the second subexpression as duplicative.