ML Duplicate Ticket Detection in Federated Systems
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In large, distributed team development environments, there is no efficient method to detect duplicate source code and tickets across different teams, projects, or systems, leading to redundant work due to inadequate communication and inconsistent naming practices.
Innovation Solution
A machine learning system is implemented to detect duplicate tickets in federated collaboration systems by tokenizing source code and ticket data, using historical ticket data as a labeled training set to train a machine learning model that identifies duplicate pairs based on digital metadata.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If manual tracking and communication methods are used in large distributed teams, then team members can independently create tickets, but duplicate tickets are created across different teams and projects
Solution Approach 1:
The patent introduces an intermediary machine learning system that acts as a mediator between ticket creation and storage. This system automatically analyzes incoming tickets, compares them against existing tickets using natural language processing and vector similarity, and identifies duplicates before they are stored in the database, thereby preventing duplicate accumulation while maintaining independent ticket creation capability
Solution Approach 2:
The system performs preliminary duplicate detection and filtering actions before tickets are fully processed and stored. By pre-analyzing ticket content, generating vector representations, and comparing against the knowledge base in advance, the system eliminates duplicates proactively rather than reactively, reducing the quantity of duplicate tickets before they enter the main system
2Reliability
If duplicate tickets are allowed to accumulate, then all team contributions are captured, but storage efficiency decreases and network bandwidth is consumed
Solution Approach 1:
The patent extracts and removes duplicate tickets from the system by implementing an automated detection and filtering mechanism. The machine learning system identifies duplicate tickets based on vector similarity and natural language analysis, then extracts only the unique, valuable tickets for storage and processing, eliminating redundant data that would consume network bandwidth and storage resources
Solution Approach 2:
The system changes the parameter of ticket representation by converting tickets into vector embeddings and using similarity thresholds for duplicate detection. By transforming the comparison parameter from exact string matching to vector space similarity, the system can efficiently identify semantic duplicates while consuming less computational resources compared to exhaustive comparison methods
3Measurement precision
If exhaustive comparison methods are used to detect duplicates, then all potential duplicates are identified, but computational complexity increases significantly
Solution Approach 1:
The patent replaces traditional mechanical exhaustive comparison methods with a machine learning-based vector similarity system. Instead of computationally intensive string matching and semantic analysis of all ticket pairs, the system uses pre-trained language models to generate vector representations and applies efficient similarity calculations, dramatically reducing computational complexity while maintaining or improving detection accuracy
Solution Approach 2:
The system performs preliminary processing by pre-computing vector embeddings for tickets and organizing them in an efficient retrieval structure. This preliminary action allows for rapid similarity comparison without requiring exhaustive real-time analysis, reducing the computational burden during actual duplicate detection while maintaining high accuracy through pre-trained model knowledge
Data Source
AI summary
A system and method for detecting duplicate tickets submitted to a ticket tracking or issue tracking system. A trained machine learning model can be leveraged to compare tokens representing ticket pairs and/or to determine whether two or more tickets are duplicate tickets, despite that content of those tickets may not be identical.