skip to main content


Search for: All records

Creators/Authors contains: "Zhou, Shurui"

Note: When clicking on a Digital Object Identifier (DOI) number, you will be taken to an external site maintained by the publisher. Some full text articles may not yet be available without a charge during the embargo (administrative interval).
What is a DOI Number?

Some links on this page may take you to non-federal websites. Their policies may differ from this site.

  1. Incorporating machine learning (ML) components into software products raises new software-engineering challenges and exacerbates existing ones. Many researchers have invested significant effort in understanding the challenges of industry practitioners working on building products with ML components, through interviews and surveys with practitioners. With the intention to aggregate and present their collective findings, we conduct a meta-summary study: We collect 50 relevant papers that together interacted with over 4758 practitioners using guidelines for systematic literature reviews. We then collected, grouped, and organized the over 500 mentions of challenges within those papers. We highlight the most commonly reported challenges and hope this meta-summary will be a useful resource for the research community to prioritize research and education in this field. 
    more » « less
    Free, publicly-accessible full text available May 1, 2024
  2. The documentation practice for machine-learned (ML) models often falls short of established practices for traditional software, which impedes model accountability and inadvertently abets inappropriate or misuse of models. Recently, model cards, a proposal for model documentation, have attracted notable attention, but their impact on the actual practice is unclear. In this work, we systematically study the model documentation in the field and investigate how to encourage more responsible and accountable documentation practice. Our analysis of publicly available model cards reveals a substantial gap between the proposal and the practice. We then design a tool named DocML aiming to (1) nudge the data scientists to comply with the model cards proposal during the model development, especially the sections related to ethics, and (2) assess and manage the documentation quality. A lab study reveals the benefit of our tool towards long-term documentation quality and accountability. 
    more » « less
    Free, publicly-accessible full text available April 19, 2024
  3. Data analysis is an exploratory, interactive, and often collaborative process. Computational notebooks have become a popular tool to support this process, among others because of their ability to interleave code, narrative text, and results. However, notebooks in practice are often criticized as hard to maintain and being of low code quality, including problems such as unused or duplicated code and out-of-order code execution. Data scientists can benefit from better tool support when maintaining and evolving notebooks. We argue that central to such tool support is identifying the structure of notebooks. We present a lightweight and accurate approach to extract notebook structure and outline several ways such structure can be used to improve maintenance tooling for notebooks, including navigation and finding alternatives. 
    more » « less
  4. The introduction of machine learning (ML) components in software projects has created the need for software engineers to collaborate with data scientists and other specialists. While collaboration can always be challenging, ML introduces additional challenges with its exploratory model development process, additional skills and knowledge needed, difficulties testing ML systems, need for continuous evolution and monitoring, and non-traditional quality requirements such as fairness and explainability. Through interviews with 45 practitioners from 28 organizations, we identified key collaboration challenges that teams face when building and deploying ML systems into production. We report on common collaboration points in the development of production ML systems for requirements, data, and integration, as well as corresponding team patterns and challenges. We find that most of these challenges center around communication, documentation, engineering, and process, and collect recommendations to address these challenges. 
    more » « less
  5. Data scientists reportedly spend 60 to 80 percent of their time in their daily routines on data wrangling, i.e. cleaning data and extracting features. However, data wrangling code is often repetitive and error-prone to write. Moreover, it is easy to introduce subtle bugs when reusing and adopting existing code, which result not in crashes but reduce model quality. To support data scientists with data wrangling, we present a technique to generate interactive documentation for data wrangling code. We use (1) program synthesis techniques to automatically summarize data transformations and (2) test case selection techniques to purposefully select representative examples from the data based on execution information collected with tailored dynamic program analysis. We demonstrate that a JupyterLab extension with our technique can provide documentation for many cells in popular notebooks and find in a user study that users with our plugin are faster and more effective at finding realistic bugs in data wrangling code. 
    more » « less
  6. The notion of forking has changed with the rise of distributed ver- sion control systems and social coding environments, like GitHub. Traditionally forking refers to splitting off an independent devel- opment branch (which we call hard forks); research on hard forks, conducted mostly in pre-GitHub days showed that hard forks were often seen critical as they may fragment a community. Today, in so- cial coding environments, open-source developers are encouraged to fork a project in order to contribute to the community (which we call social forks), which may have also influenced perceptions and practices around hard forks. To revisit hard forks, we identify, study, and classify 15,306 hard forks on GitHub and interview 18 owners of hard forks or forked repositories. We find that, among others, hard forks often evolve out of social forks rather than being planned deliberately and that perception about hard forks have indeed changed dramatically, seeing them often as a positive non- competitive alternative to the original project. 
    more » « less
  7. With the emergence of social coding platforms, collaboration has become a key and dynamic aspect to the success of software projects. In such platforms, developers have to collaborate and deal with issues of collaboration in open-source software development. Although collaboration is challenging, collaborative development produces better software systems than any developer could produce alone. Several approaches have investigated collaboration challenges, for instance, by proposing or evaluating models and tools to support collaborative work. Despite the undeniable importance of the existing efforts in this direction, there are few works on collaboration from perspectives of developers. In this work, we aim to investigate the perceptions of open-source software developers on collaborations, such as motivations, techniques, and tools to support global, productive, and collaborative development. Following an ad hoc literature review, an exploratory interview study with 12 open-source software developers from GitHub, our novel approach for this problem also relies on an extensive survey with 121 developers to confirm or refute the interview results. We found different collaborative contributions, such as managing change requests. Besides, we observed that most collaborators prefer to collaborate with the core team instead of their peers. We also found that most collaboration happens in software development (60%) and maintenance (47%) tasks. Furthermore, despite personal preferences to work independently, developers still consider collaborating with others in specific task categories, for instance, software development. Finally, developers also expressed the importance of the social coding platforms, such as GitHub, to support maintainers, and contributors in making decisions and developing tasks of the projects. Therefore, these findings may help project leaders optimize the collaborations among developers and reduce entry barriers. Moreover, these findings may support the project collaborators in understanding the collaboration process and engaging others in the project. 
    more » « less
  8. Forking and pull requests have been widely used in open-source communities as a uniform development and contribution mechanism, giving developers the flexibility to modify their own fork without affecting others before attempting to contribute back. However, not all projects use forks efficiently; many experience lost and duplicate contributions and fragmented communities. In this paper, we explore how open-source projects on GitHub differ with regard to forking inefficiencies. First, we observed that different communities experience these inefficiencies to widely different degrees and interviewed practitioners to understand why. Then, using multiple regression modeling, we analyzed which context factors correlate with fewer inefficiencies.We found that better modularity and centralized management are associated with more contributions and a higher fraction of accepted pull requests, suggesting specific best practices that project maintainers can adopt to reduce forking-related inefficiencies in their communities. 
    more » « less
  9. Fork-based development is popular and easy to use, but makes it difficult to maintain an overview of the whole community when the number of forks increases. This may lead to redundant development where multiple developers are solving the same problem in parallel without being aware of each other. Redundant development wastes effort for both maintainers and developers. In this paper, we designed an approach to identify redundant code changes in forks as early as possible by extracting clues indicating similarities between code changes, and building a machine learning model to predict redundancies. We evaluated the effectiveness from both the maintainer's and the developer's perspectives. The result shows that we achieve 57-83% precision for detecting duplicate code changes from maintainer's perspective, and we could save developers' effort of 1.9-3.0 commits on average. Also, we show that our approach significantly outperforms existing state-of-art. 
    more » « less