文档介绍：
基于词性标注的特征定位方法#
张希远1,李宏伟2,3,姜诗海1,彭鑫2,赵文耘2**
(1. 复旦大学软件学院,上海
201203;
5
2. 复旦大学计算机科学技术学院,上海
201203;
3. 江西师范大学计算机信息工程学院,南昌 330022)
摘要:特征与相关实现代码之间关系的逆向恢复被称为特征定位。现有特征定位技术主要依
赖于分析动态执行情况或程序语法结构,无法充分利用语义关联信息,因此在结果的完整性
和正确性上存在不足。在使用基于信息检索技术的基础上,本文提出了引入词性标注,结合
10
程序结构信息抽取和语义分析的特征定位方法。实验结果表明在文本匹配的基础上结合短语
匹配能在一定程度上提高查全率。特征定位结果的查全率取决于该特征的自然语言描述与其
在代码中表述的差异。
关键词:软件工程;特征定位;词性标注;ard 系数
中图分类号:
15
A Lexical Category Labeling Method for Feature Location
ZANG Xiyuan1, LI Hongwei2,3, JIANG Shihai1, PENG Xin2, ZHAO Wenyun2
(1. School of Software, Fudan University, Shanghai 201203;
2. Department puter Science, Fudan University, Shanghai 201203;
20
25
30
35
3. School puter Information and Engineering, Jiangxi Normal University, Nanchang
330022)
Abstract: Feature location is the activity of reconverying relationship between feature and.
implementation. Existing feature location technologies mainly depends on examining a software
system's execution trace or analyzing program's structural information. These methods unable to
take full advantage of semantic information, so the results lack pletness and correctness.
This paper attempt to locate feature based on lexical category labeling of natural Language, With
IR technology as the foundation. Make use of lexical category labeling to locate feature by
combining the code information extraction and semantic an