| 本文已被：浏览 2601次 下载 4834次 |
码上扫一扫！ |
|
|
| 基于代码结构知识的软件文档语义搜索方法 |
|
林泽琦1,2, 邹艳珍1,2,3, 赵俊峰1,2,3, 曹英魁1,2, 谢冰1,2
|
|
1.高可信软件技术教育部重点实验室(北京大学), 北京 100871;2.北京大学 信息科学技术学院, 北京 100871;3.北京大学(天津滨海)新一代信息技术研究院, 天津 300450
|
|
| 摘要: |
| 自然语言文本形式的文档是软件项目的重要组成部分.如何帮助开发者在大量文档中进行高效、准确的信息定位，是软件复用领域中的一个重要研究问题.提出了一种基于代码结构知识的软件文档语义搜索方法.该方法从软件项目的源代码中解析出代码结构图，并以此作为领域特定的知识来帮助机器理解自然语言文本的语义.这一语义信息与信息检索技术相结合，从而实现了对软件文档的语义检索.在StackOverflow问答文档数据集上的实验表明，与多种文本检索方法相比，该方法在平均准确率（mean average precision，简称MAP）上可以取得至少13.77%的提升. |
| 关键词: 软件复用 自然语言文本 代码结构知识 信息检索 语义搜索 |
| DOI：10.13328/j.cnki.jos.005609 |
| 分类号:TP311 |
| 基金项目:国家重点研发计划（2016YFB1000801）；国家杰出青年科学基金（61525201） |
|
| Software Text Semantic Search Approach Based on Code Structure Knowledge |
|
LIN Ze-Qi1,2, ZOU Yan-Zhen1,2,3, ZHAO Jun-Feng1,2,3, CAO Ying-Kui1,2, XIE Bing1,2
|
|
1.Key Laboratory of High Confidence Software Technologies(Peking University), Ministry of Education, Beijing 100871, China;2.School of Electronics Engineering and Computer Science, Peking University, Beijing 100871, China;3.Peking University Information Technology Institute(Tianjin Binhai), Tianjin 300450, China
|
| Abstract: |
| Natural language text is a common form of knowledge representation in various software artifacts. During the practice of software reuse, software developers usually need to search the large amount of textual resource. This paper presents a software text semantic search approach based on code structure knowledge. This approach extracts a code structure graph from software source code and leverages it as a domain-specific knowledge base to analyze the semantic meanings of natural language texts. The semantic information is combined with information retrieval technology to re-rank text search results semantically. Experimental results on StackOverflow dataset show that this approach achieves at least 13.77% improvement in mean average precision (MAP) comparing to several text retrieval approaches. |
| Key words: software reuse natural language text code structure knowledge information retrieval semantic search |
