SHF: Medium: Collaborative Research: Semantically-Enhanced Software Traceability for Supporting Human-Centric Tasks

SHF：媒介：协作研究：语义增强的软件可追溯性，支持以人为本的任务

• 批准号: 1901059
• 负责人: Jane Huang
• 金额: $ 101.9万
• 依托单位: University of Notre Dame
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 2019
• 资助国家: 美国
• 起止时间: 2019-08-15 至 2024-07-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1901059&HistoricalAwards=false
• 关键词: SHF; Medium; Collaborative; Research; Semantically

Achieving accurate, complete, trustworthy and usable traceability across software-intensive systems can be extremely beneficial as the underlying network of traceability links can be used to answer diverse questions about the software system and its development process. However, in practice, many software projects suffer from inadequate or inaccurate traceability due to the cost and difficulty of manually creating and maintaining trace links. This research will develop a holistic, interactive tracing environment, which incorporates diverse algorithmic solutions for dynamically generating trace links, visualizing the results, and guiding the user through the interactive process of using the results to support diverse Software Engineering tasks. Automating the creation and maintenance of accurate traceability links offers significant potential for industrial impact. For example, traceability is required by certifying bodies in safety-critical domains and can help in the construction and delivery of high quality, competitive, timely products. The cross-disciplinary nature of the team will introduce new opportunities for software engineers, data scientists, and human-computer interaction experts to collaborate in addressing open Software Engineering challenges and will provide research opportunities for diverse and underrepresented students.The research will explore challenging problems at the intersection of software engineering, semantic text mining, and visualization. It will directly address one of the prominent causes of trace-link inaccuracy caused by the inability of current algorithms to reason over deep semantics of underlying software artifacts such as requirements, design, and code. First, the researchers will investigate semantically enhanced traceability algorithms that generate trace links, even in the absence of shared textual representations. Second, the work will develop a holistic tracing solution that dynamically configures a trace engine to leverage diverse tracing techniques such as semantic traceability, trace link evolution, and other existing techniques. Finally, given a diverse set of trace links with different degrees of accuracy, the research team will design, develop, and publicly release a novel, interactive, visual interface that enables users to understand the provenance and trustworthiness of each link while providing a clear rationale for trace query results.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

在软件密集型系统中实现准确，完整，值得信赖和可用的可追溯性可能非常有益，因为可追溯性链接的基础网络可用于回答有关软件系统及其开发过程的各种问题。 但是，实际上，由于手动创建和维护跟踪链接的成本和困难，许多软件项目的可追溯性不足或不正确。 这项研究将开发一个整体，交互式的跟踪环境，该环境结合了多种算法解决方案，以动态生成跟踪链接，可视化结果并通过使用结果来支持各种软件工程任务的交互过程。自动化准确的可追溯性链接的创建和维护为工业影响带来了巨大的潜力。例如，通过在安全至关重要的域中认证机构需要可追溯性，并可以帮助建造和交付高质量，竞争性，及时的产品。 团队的跨学科性质将为软件工程师，数据科学家和人力计算机互动专家提供新的机会，以协作解决开放软件工程挑战，并为多元化和代表性不足的学生提供研究机会。该研究将探索在软件工程，语义文本，语义文本矿业和可视化的相互作用方面的挑战性问题。 它将直接解决由于当前算法无法对基础软件文物（例如需求，设计和代码）进行深层语义的推理，这是由于当前算法无法推理了痕量链接不准确的主要原因之一。 首先，研究人员将研究即使没有共享的文本表示，即使在没有共享的文本表示的情况下，也会研究产生跟踪链接的可追溯性算法。其次，这项工作将开发一种整体跟踪解决方案，该解决方案将动态配置痕量引擎以利用多样化的跟踪技术，例如语义可追溯性，跟踪链路演变和其他现有技术。最后，考虑到具有不同程度准确性的各种痕量链接，研究团队将设计，开发和公开发布一种新颖的，互动的，视觉界面，使用户能够理解每个链接的出处和信任度，同时为痕迹询问结果提供了明确的基本原理，以反映NSF的法定任务，并通过评估了基础，这一奖项是通过评估的范围来进行的。

项目成果

Traceability Transformed: Generating more Accurate Links with Pre-Trained BERT Models

• DOI: 10.1109/icse43902.2021.00040
• 发表时间: 2021-01-01
• 期刊: 2021 IEEE/ACM 43RD INTERNATIONAL CONFERENCE ON SOFTWARE ENGINEERING (ICSE 2021)
• 作者: Lin, Jinfeng;Liu, Yalin;Cleland-Huang, Jane
• 通讯作者: Cleland-Huang, Jane

Generate rather than Retrieve: Large Language Models are Strong Context Generators

• DOI: 10.48550/arxiv.2209.10063
• 发表时间: 2022-09
• 期刊: ArXiv
• 作者: W. Yu;Dan Iter;Shuohang Wang;Yichong Xu;Mingxuan Ju;Soumya Sanyal;Chenguang Zhu;Michael Zeng;Meng Jiang

Leveraging Intermediate Artifacts to Improve Automated Trace Link Retrieval

利用中间工件来改进自动跟踪链接检索

• DOI: 10.1109/icsme52107.2021.00014
• 发表时间: 2021
• 期刊: IEEE International Conference on Software Maintenance and Evolution
• 作者: Rodriguez, Alberto D.;Cleland-Huang, Jane;Falessi, Davide
• 通讯作者: Falessi, Davide

Dict-BERT: Enhancing Language Model Pre-training with Dictionary

• DOI: 10.18653/v1/2022.findings-acl.150
• 发表时间: 2021-10
• 作者: W. Yu;Chenguang Zhu;Yuwei Fang;Donghan Yu;Shuohang Wang;Yichong Xu;Michael Zeng;Meng Jiang

Exploring Contrast Consistency of Open-Domain Question Answering Systems on Minimally Edited Questions

探索开放域问答系统对最少编辑问题的对比度一致性

• 发表时间: 2023
• 期刊: Transactions of the Association for Computational Linguistics
• 影响因子: 10.9
• 作者: Zhang, Zhihan;Yu, Wenhao;Ning, Zheng;Ju, Mingxuan;Jiang, Meng
• 通讯作者: Jiang, Meng