Vector Embedding for Semantic Issue Similarity Detection

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing data processing systems lack efficient mechanisms to identify and retrieve electronic content items similar in meaning, particularly in collaborative software applications like issue tracking systems, where semantic similarity analysis is crucial for preventing duplication and detecting incidents.

Innovation Solution

The system employs machine learning and artificial intelligence through word-embedding models to convert issue titles and descriptions into vector numbers, allowing for the computation of distances between them, thereby identifying semantically similar issues and detecting incidents by determining threshold similarities in newly created issues.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If manual review of issues is performed to identify similar content items, then accuracy of similarity identification can be maintained, but time consumption and operational efficiency deteriorate

Engineering Contradiction:

Improveaccuracy of similarity identificationVSAvoidtime consumption

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The patent replaces manual mechanical review processes with automated machine learning systems. Specifically, it uses natural language processing models to encode issue descriptions into vector representations and computes similarity metrics automatically, eliminating the need for human reviewers to manually compare issues while maintaining or improving accuracy through consistent algorithmic application.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The patent introduces vector embeddings as an intermediary representation between raw issue text and similarity assessment. By converting textual descriptions into numerical vector spaces where semantic meaning is preserved, the system enables efficient computational comparison without requiring direct text analysis, thus reducing time consumption while maintaining measurement precision.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Productivity

If automated similarity detection is implemented using machine learning, then operational efficiency is improved, but system complexity increases

Engineering Contradiction:

Improveoperational efficiencyVSAvoidsystem complexity

Core Design Contradiction:

ProductivityVSDevice complexity

Solution Approach 1:

The patent employs a universal sentence encoding model that can process diverse issue descriptions across different domains and contexts. This single multi-functional model handles various types of technical support issues without requiring separate specialized systems, thereby improving operational efficiency while limiting the growth of system complexity through consolidation.

Inventive Principle:

Principle #6Universality (Multi-functionality)

Solution Approach 2:

The patent creates simplified vector representations (copies) of complex textual issue descriptions. These vector copies capture the essential semantic information in a compact numerical form that is easy to compute and compare, enabling efficient automated processing without requiring the system to handle the full complexity of natural language text directly.

Inventive Principle:

Principle #26Copying

3Reliability

If comprehensive issue analysis is performed to detect incidents, then reliability of incident detection is improved, but computational resources consumed increase

Engineering Contradiction:

Improvereliability of incident detectionVSAvoidcomputational resources

Core Design Contradiction:

ReliabilityVSUse of energy by moving object

Solution Approach 1:

The patent implements a two-stage analysis approach where issues are first screened using efficient vector similarity computations, and only potentially relevant issues undergo more comprehensive analysis. This partial action strategy maintains high reliability by thoroughly analyzing only the necessary subset of issues while reducing overall computational resource consumption by avoiding exhaustive analysis of all issues.

Inventive Principle:

Principle #16Partial or excessive action

Solution Approach 2:

The patent transforms issue descriptions from textual form to vector representations, changing the parameter space from discrete text tokens to continuous numerical vectors. This parameter transformation enables more efficient computational operations such as distance calculations and similarity measurements, reducing computational resource requirements while maintaining or improving the reliability of incident detection through better numerical stability and faster processing.

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentUS11544122B2Systems and methods for identifying similar electronic content items

Publication Date: 2023.01.03 ATLASSIAN PTY LTD

AI summary

Systems, methods and non-transitory computer readable media for detecting incidents are disclosed. The method includes receiving a primary issue creation event record for a primary issue, the event record including a description of the primary issue, and encoding the primary issue into a primary vector number based on the description of the primary issue. The method further includes identifying candidate issues and retrieving vector numbers of the identified candidate issues, computing distances between the primary vector number and each of the candidate vector numbers, and determining whether incident criteria is met based on the computed distances. In addition, the method includes determining that an incident has occurred upon determining that the incident criteria is met and generating an alert.