Application programming interface (API) is an important form of software reuse. API documentations, such as API specifications, tutorials, and online forums, are valuable learning resources for reusing the APIs. In recent years, many data‐driven API documentation mining (ADM) methods have been proposed. These methods mine API documentations and return API‐related information to help developers better understand and reuse APIs. These methods treat documentations as unstructured data and apply various data mining techniques to analyze the documentation data. Currently, there is no comprehensive review of the data‐driven approach to API documentation mining. This review aims to fill in this gap by analyzing and discussing the state of the art ADM papers. We survey 32 representative papers published in prominent software engineering journals and conferences in recent 5 years (January 2014–July 2019). We analyze their mining tasks, mined data, problems, data mining techniques, and evaluation metrics. Based on the survey results, we point out research challenges and future research directions in this area.

中文翻译：

---

应用程序编程接口文档挖掘的数据驱动方法：回顾

应用程序编程接口（API）是软件重用的一种重要形式。API文档（例如API规范，教程和在线论坛）是重用API的宝贵学习资源。近年来，已经提出了许多数据驱动的API文档挖掘（ADM）方法。这些方法挖掘API文档并返回与API相关的信息，以帮助开发人员更好地理解和重用API。这些方法将文档视为非结构化数据，并应用各种数据挖掘技术来分析文档数据。当前，没有对API文档挖掘的数据驱动方法进行全面审查。这篇综述旨在通过分析和讨论最新的ADM论文来填补这一空白。我们调查了最近5年（2014年1月至2019年7月）在著名软件工程期刊和会议上发表的32篇代表性论文。我们分析他们的挖掘任务，挖掘的数据，问题，数据挖掘技术和评估指标。根据调查结果，我们指出了该领域的研究挑战和未来的研究方向。