Deep learning-based automatic generation method for single sentence abstract defect report title

Abstract

The invention provides a deep learning-based automatic generation method for a single sentence abstract defect report title. The method comprises the following steps: acquiring an open source defect report sample as an original data set, constructing three discrimination models to evaluate the sample in the original data set, and selecting a qualified sample to construct a formal data set; introducing a plurality of artificially named vocabulary types and constructing a corresponding regular expression for extracting and positioning artificially named vocabularies in the main body content of the formal data set sample; inserting type marks in front of and behind each artificial named vocabulary, and performing word segmentation and minimization processing on titles of the samples and the marked main content to construct a training data set; building a coding and decoding recurrent neural network model with a copying mechanism, and training on the training data set to obtain an optimized model; and inputting the defect report main body content of the title to be drafted into the optimized model, so that the model can automatically generate the corresponding title. According to the invention, the title writing quality and efficiency of the user are improved.

Description

technical field [0001] The invention relates to the field of computer technology, in particular to a method for automatically generating a single-sentence summary defect report title based on deep learning. Background technique [0002] Software defect reports play a vital role in software development and are included in a large number of software repositories. Defect reports with high-quality content help to understand, reproduce, locate and fix software defects. As one of the required contents of a software defect report, a high-quality defect report title can convey the core idea of ​​the specific details of the defect report, help project maintainers quickly understand the defect report, and assign appropriate labels to it more effectively. prioritization, or develop an appropriate defect triage strategy. However, due to inconsistent levels, limited writing time, etc., it is difficult to guarantee the quality of manually written defect report titles in the course of pr...

AI Technical Summary

Problems solved by technology

[0007] There is no ready-made high-quality data set, and the quality of data obtained from open source channels varies and cannot be directly used as a data set

The neural network model based on big data learning always needs high-quality data samples that meet the task requirements for training, otherwise problems such as low training efficiency or wrong learning direction may occur during model training

The unfiltered data set obtained directly from the open source community may contain a large number of invalid defect report samples with poor title quality and non-single-sentence summary, which cannot be directly used as a training data set

[0008] Cannot effectively deal with artificially named words with low word frequency. Artificially named words such as identifiers and version numbers are often specific to defect reports and their software warehouses. Different projects usually contain different artificially named words. Therefore, defect reports There are inexhaustible spelling forms of artificially named words in , and the word frequency of each artificially named word is relatively low

However, since the neural summarization model needs to learn, understand and generate target word examples by repeatedly adjusting word embedding vectors and other related parameters recorded in the fixed word example table of the model, the low-frequency words in the corpus are usually not well understood. processed by the model

Images