SHF: SMALL: Automatically Supporting Developer Learning in Context

努力创建和发展大型软件系统的软件开发人员发现他们经常需要学习。他们加入了一个新的项目，在那里他们需要学习一个新的软件系统和开发团队？的文化。 他们经常需要学习新的应用程序编程接口（API）或现有API的更新组件。该项目解决了对有用的自动化支持的需求，以帮助软件开发人员学习API以及如何有效地使用它们来解决手头的任务。所设想的系统执行基本的自动化分析的服务器上，从用户的信息？的上下文和当前维护或进化任务，并在执行任务时发回学习数据。 有了更有效的学习方法，软件开发可以更有效率，并提供更可靠，更高质量的软件，这对越来越依赖软件的社会产生了广泛的影响。这个项目将通过解决三个主要挑战来促进最先进的技术，将这种自动化的开发人员学习支持投入实际使用。首先，我们将开发分析技术，以弥合单个语句和整个方法级分析之间的差距，从而在多语句、算法步骤级从源代码中自动提取、描述和概括信息。其次，我们将开发工具，自动识别，提取和分类不同类型的信息，如事实，（积极和消极）意见/建议，以及混合文本代码工件（如电子邮件，问答论坛和其他开发人员通信）中的使用信息。第三，我们将开发工具来自动识别开发人员的相关上下文，并识别该上下文的相关学习内容。新的分析方法，由此产生的工具，数据集和实验基础设施开发的项目将被释放，这将使其他研究人员和从业人员建立在该项目？的结果，并将最终推进软件工程领域内的知识和理解。

Software developers striving to create and evolve large software systems find themselves frequently in need of learning. They join a new project where they need to learn a new software system and the development team?s culture. They frequently need to learn new Application Programming Interfaces (APIs), or newer components of existing APIs. The project addresses the need for useful automated support for helping software developers learn APIs and how to use them effectively solve the task at hand. The envisioned system performs the underlying automated analysis on a server which takes information from the user?s context and current maintenance or evolution task, as allowed by the user, and sends back learning nuggets as they perform their tasks. With more effective ways to learn, software development could be more efficient and provide more reliable, higher quality software, which has a broad impact on society which depends increasingly on software.This project will contribute to the state of the art by tackling three major challenges to bring this kind of automated support for developer learning into practical use. First, we will develop analysis techniques to bridge the gap between single statement and whole method level analyses for automatic extraction, description, and generalization of information from source code at the multi-statement, algorithm-step level. Second, we will develop tools that automatically identify, extract and categorize different kinds of information such as facts, (positive and negative) opinions/advice, and usage information in mixed text-code artifacts such as emails, question-answer forums, and other developer communications. Third, we will develop tools to automatically identify the relevant context of the developer and identify the relevant learning nuggets for that context. The novel approach to analyses, the resulting tools, data sets, and experimental infrastructure developed within the project will be released, which will enable other researchers and practitioners to build on the project?s results and will ultimately advance knowledge and understanding within the field of Software Engineering.