SHF: Small: Enabling Software Engineering Virtual Assistant Technology

SHF：小型：启用软件工程虚拟助理技术

• 批准号: 1717607
• 负责人: Collin McMillan
• 金额: $ 40.72万
• 依托单位: University of Notre Dame
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2017
• 资助国家: 美国
• 起止时间: 2017-09-01 至 2021-08-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1717607&HistoricalAwards=false
• 关键词: SHF; Small; Enabling; Software; Engineering

The objective of this research project is to address key barriers in adapting natural language processing techniques to problems in creating virtual assistants for software engineering. Virtual assistants such as Siri, Cortana, and Alexa are claiming an increasing role in computing for everyday tasks, but multiple barriers prevent existing virtual assistant technology from being applied to software engineering tasks. The long-term goal of the project is that virtual assistants will improve productivity for software engineers.This proposal targets two of those barriers: 1) conversation analysis and modeling, and 2) reference expression generation. The first of these problems, in a nutshell, is that experiments in natural language modeling conversations tend to cover topics with similar outcomes, while conversations about software may have a much wider range of possible outcomes. The second problem is that much research in natural language processing is focused on how humans refer to physical objects that have attributes that are universally preferred while in contrast, software artifacts tend not to have measurable attributes that people use as descriptions. The chief broader impact is an application to assistive technology for persons who are visually impaired. Virtual assistants have the potential to alleviate barriers-to-entry into computing careers faced by visually impaired persons by creating a voice interface for answering software development questions.

该研究项目的目的是解决将自然语言处理技术调整到为软件工程创建虚拟助手时的问题的关键障碍。 Siri，Cortana和Alexa等虚拟助手声称在日常任务中的计算中发挥了越来越多的作用，但是多个障碍阻止了现有的虚拟助手技术被应用于软件工程任务。 该项目的长期目标是虚拟助手将提高软件工程师的生产率。该提案针对其中两个障碍：1）对话分析和建模，以及2）参考表达生成。 简而言之，其中的第一个问题是，自然语言建模对话的实验倾向于涵盖具有相似结果的主题，而有关软件的对话可能会产生更多可能的结果。 第二个问题是，自然语言处理中的许多研究集中在人类如何提及具有普遍优选的属性的物理对象上，而相比之下，软件伪像往往没有可测量的属性，而这些属性是人们用作描述的可测量属性。 最广泛的影响是针对视力障碍者的辅助技术的应用。 虚拟助手有可能通过创建一个用于回答软件开发问题的语音界面来减轻视力障碍者面临的计算职业的障碍。

项目成果

Ensemble Models for Neural Source Code Summarization of Subroutines

• DOI: 10.26226/morressier.613b5418842293c031b5b62e
• 发表时间: 2021-07
• 期刊: 2021 IEEE International Conference on Software Maintenance and Evolution (ICSME)
• 作者: Alexander LeClair;Aakash Bansal;Collin McMillan

Dialogue Management for Interactive API Search

• DOI: 10.26226/morressier.613b5418842293c031b5b5e8
• 发表时间: 2021-07
• 期刊: 2021 IEEE International Conference on Software Maintenance and Evolution (ICSME)
• 作者: Zachary Eberhart;Collin McMillan

A Wizard of Oz Study Simulating API Usage Dialogues With a Virtual Assistant

• DOI: 10.1109/tse.2020.3040935
• 发表时间: 2021-06
• 期刊: IEEE Transactions on Software Engineering
• 影响因子: 7.4
• 作者: Zachary Eberhart;Aakash Bansal;Collin McMillan

A Neural Question Answering System for Basic Questions about Subroutines

• DOI: 10.1109/saner50967.2021.00015
• 发表时间: 2021-01
• 期刊: 2021 IEEE International Conference on Software Analysis, Evolution and Reengineering (SANER)
• 作者: Aakash Bansal;Zachary Eberhart;Lingfei Wu;Collin McMillan

Action Word Prediction for Neural Source Code Summarization

• DOI: 10.1109/saner50967.2021.00038
• 发表时间: 2021-01
• 期刊: 2021 IEEE International Conference on Software Analysis, Evolution and Reengineering (SANER)
• 作者: S. Haque;Aakash Bansal;Lingfei Wu;Collin McMillan