一种基于命名实体识别的需求跟踪方法

针对基于文本的需求跟踪方法严重依赖文本质量的问题,提出了一种利用命名实体识别技术标注制品文档关键词的需求跟踪方法。该方法通过代码实体上下文构建命名实体识别模型,解决了抽象语法树和正则表达式无法解析非源代码形式的软件制品的问题。利用命名实体识别模型标识出软件制品中的代码实体之后,方法将软件制品转换为文档集合并进行语义聚类,最后再通过映射算法创建制品间的需求跟踪关系。实验结果表明,与基于所有词项和基于高权重词项的需求跟踪方法相比,该方法能够有效提高需求跟踪结果的质量。

Recovering traceability links using named entity recognition

Aiming at the problem that requirement traceability approaches based on textual information were rely heavily on the quality of the text, this paper proposed a traceability approach utilized named entity recognition technology to identify key words in software artefacts. Firstly, the proposed method constructed a named entity recognition model through the context of code entity, which solved the issue that abstract syntax tree and the regular expression was not able to parse non-source form software artefacts. After that, the proposed method transformed software artefacts to document set, and then carried out a semantic clustering process to cluster documents. Finally, the proposed method created trace links between software artefacts using the mapping algorithm. The experimental results showed that comparing with those traceability approaches based on the all terms and high weight terms, our proposal was able to effectively improve the quality of requirement tracing results.