Automatic Requirements Classification Based on Graph Attention Network

Abstract

Requirements classification is a significant task for requirements engineering, which is timeconsuming and challenging. The traditional requirements classification models usually rely on manual preprocessing and have poor generalization capability. Moreover, these traditional models ignore the sentence structure and syntactic information in requirements. To address these problems, we propose an automatic requirements classification based BERT and graph attention network (GAT), called DBGAT. We construct dependency parse trees and then utilize the GAT for mining the implicit structure feature and syntactic feature of requirements. In addition, we introduce BERT to improve the generalization ability of the model. Experimental results of the PROMISE datasets demonstrate that our proposed DBGAT significantly outperforms existing state-of-the-art methods. Moreover, we investigate the impact of graph construction methods on non-functional requirements classification. DBGAT achieved the best classification results on both seen (F1-scores of up to 91%) and unseen projects (F1-scores of up to 88%), further demonstrating the strong generalization ability.

Detecting Security Requirements in GitHub Issues -Novel Dataset and SmallBERT-based model

“…This involves binary classification, often using traditional machine-learning models. Research has explored classifying requirements from software specifications, as exemplified by works such as [10]- [13]. Security, a critical non-functional requirement, is crucial for application integrity and was examined in classification tasks by [14]- [16].…”

“…Furthermore, Gang Li and Chengpeng Zheng explored multiple approaches for classifying functional and non-functional requirements, with the combination of Generative Adversarial Network (GAN) and BERT demonstrating the best performance, achieving an impressive F1 score of 0.91 [13]. Gouri Deshpande's study [18] compared the performance of two models, BERT and Random Forest, for classifying dependent requirements.…”

“…PROMISE [19], a widely employed dataset, encompasses 625 requirement statements, including 255 functional and 370 non-functional requirements. It has been utilized in numerous studies, e.g., [11]- [13], [17], [20], with further categorization performed in [17] based on quality attributes.…”

Learning Machine Translation with Linguistic Interpretation

“…performance but also improves its interpretability, a crucial aspect in understanding and improving NMT systems [11,12]. The integration of GAT into NMT poses a compelling question: Can the fusion of explicit syntactic knowledge via GAT with the deep, implicit linguistic understanding of BERT lead to a significant leap in translation quality?…”