In a software development life cycle, software requirements specifications (SRS) written in an incomprehensible language might hinder the success of the project in later stages. In such cases, the subjective and ambiguous nature of the natural languages can be considered as a cause for the failure of the final product. Redundancy and/or controversial information in the SRS documents might also result in additional costs and time loss, reducing the overall efficiency of the project. With the recent advances in machine learning, there is an increased effort to develop automated solutions for a seamless SRS design. However, most vanilla machine learning approaches ignore the semantics of the software artifacts or integrating domain-specific knowledge into the underlying natural language processing tasks, and therefore tend to generate inaccurate results. With such concerns in mind, we consider a transfer learning approach in our study, which is based on an existing pre-trained language model called DistilBERT. We specifically examine the DistilBERT's ability in multi-class text classification on SRS data using various finetuning methods, and compare its performance with other deep learning methods such as LSTM and BiLSTM. We test the performance of these models using two datasets: DOORS Next Generation dataset and PROMISE-NFR dataset. Our numerical results demonstrate that DistilBERT perform well for various text classification tasks over the SRS datasets and shows significant promise to be used for automating the software development processes.
Background : Developers spend a significant amount of time and efforts to localize bugs. In the literature, many researchers proposed state-of-the-art bug localization models to help developers localize bugs easily. The practitioners, on the other hand, expect a bug localization tool to meet certain criteria, such as trustworthiness, scalability, and efficiency. The current models are not capable of meeting these criteria, making it harder to adopt these models in practice. Recently, deep-learning-based bug localization models have been proposed in the literature, which show a better performance than the state-of-the-art models.Aim: In this research, we would like to investigate whether deep learning models meet the expectations of practitioners or not.Method : We constructed a Convolution Neural Network and a Simple Logistic model to examine their effectiveness in localizing bugs. We train these models on five open source projects written in Java and compare their performance with the performance of other state-of-the-art models trained on these datasets.Results: Our experiments show that although the deep learning models perform better than classic machine learning models, they meet the adoption criteria set by the practitioners only partially.Conclusions: This work provides evidence that the practitioners should be cautious while using the current state of the art models for production-level use-cases. It also highlights the need for standardization of performance benchmarks to ensure that bug localization models are assessed equitably and realistically.
Early detection of Alzheimer's disease (AD) is significant for identifying of better treatment plans for the patients as the AD is not curable. On the other hand, lack of interpretability for the high performing prediction models might prevent incorporation of such models in clinical usage for AD detection. Accordingly, it is important to develop highly interpretable models which can create trust towards the prediction models by showing the factors that contribute to the models' decisions. In this paper, we use ProtoPNet architecture in combination with popular pretrained deep learning models to add interpretability to the AD classifications over MRI scans from ADNI and OASIS datasets. We find that the ProtoPNet model with DenseNet121 architecture can reach 90 percent accuracy while providing explanatory illustrations of the model's reasonings for the generated predictions. We also note that, in most cases, the performances of the ProtoPNet models are slightly inferior to their black-box counterparts, however, their ability to provide reasoning and transparency in the prediction generation process can contribute to higher adoption of the prediction models in clinical practice.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.