Neural Network Bug Report Duplicate Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing duplicate bug report detection systems are inefficient as they require structured data formatting, limiting their accuracy and applicability, and often waste resources due to the need for costly rework and computing power in software development projects.
Innovation Solution
The implementation of a machine learning model using different types of neural networks (such as single-layer, bi-LSTM, and convolutional neural networks) for processing various description types within bug reports, allowing for duplicate detection without structured formatting, and enabling transfer of training data between projects.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional duplicate bug report detection systems are used, then structured data formatting is required, but this limits accuracy and applicability to free-form text
Solution Approach 1:
The patent replaces traditional mechanical text processing systems with neural network-based deep learning models. Specifically, it uses character-level CNNs and bidirectional LSTMs to automatically learn features from raw text without requiring manual structuring or formatting, thereby achieving high detection accuracy on free-form bug reports while eliminating the need for structured data input
Solution Approach 2:
The patent changes the fundamental parameters of text processing by shifting from word-level to character-level analysis. This parameter change enables the system to handle unstructured free-form text effectively, as character-level processing captures subtle patterns and variations that word-level processing would miss, thereby improving both accuracy and adaptability
2Loss of energy
If duplicate bug report detection is performed manually, then resource waste is reduced, but this increases time consumption and computational costs
Solution Approach 1:
The patent implements preliminary action by pre-training character embedding matrices and training neural network models on large corpora of bug reports before actual detection. This pre-processing enables the system to quickly classify new bug reports as duplicates or non-duplicates with high accuracy, significantly reducing both time consumption and computational resources during operational phase compared to manual review
Solution Approach 2:
The patent uses copying by creating dense vector representations (embeddings) of bug reports that capture their semantic meaning. These vector copies can be efficiently compared using similarity metrics like cosine similarity, enabling rapid duplicate detection without processing the full text each time, thereby reducing computational overhead and time
3Measurement precision
If training data is project-specific, then detection accuracy for that project is improved, but this reduces applicability across multiple projects
Solution Approach 1:
The patent achieves universality by designing a project-agnostic character-level processing pipeline. The character embedding layer and neural network architecture work universally across different projects and domains. The system can be applied to any bug report corpus without requiring project-specific customization, making it multi-functional and broadly applicable while maintaining accuracy through its ability to learn project-specific patterns from the input data
Solution Approach 2:
The patent applies segmentation by breaking down bug reports into individual characters rather than relying on project-specific tokenization or word segmentation rules. This fine-grained segmentation approach is universal across languages and domains, enabling the same model architecture to effectively process bug reports from different projects without requiring project-specific preprocessing adjustments
Data Source
AI summary
A device may receive information associated with first and second bug reports to be classified as duplicate or non-duplicate bug reports. The device may identify first and second descriptions associated with the first and second bug reports, respectively. The first and second descriptions may be different descriptions having a shared description type. The device may identify a neural network for encoding the first and second descriptions, based on the shared description type. The device may encode the first description into a first vector using the neural network, and may encode the second description into a second vector using the neural network. The device may classify the first and second bug reports as duplicate or non-duplicate bug reports based on the first vector and the second vector. The device may perform an action based on classifying the first and second bug reports as duplicate or non-duplicate bug reports.