CAJ | 학술논문

文章提出了一种提取中文软件文档与源代码间的关联关系的方法,根据中文软件文档和源代码的一些特征,在潜在语义索引模型的基础上使用了3种策略:引入项目数据词典辅助中文分词和中英文的翻译、将文档按类型分层以实现反馈,调整代码中的特征项的权值.实验结果表明,同时使用3种策略可以在提取阀值C相同的情况下,提高查全率4%～28%,在查全率不变的情况下,提高查准率8%～30%.
문장제출료일충제취중문연건문당여원대마간적관련관계적방법,근거중문연건문당화원대마적일사특정,재잠재어의색인모형적기출상사용료3충책략:인입항목수거사전보조중문분사화중영문적번역、장문당안류형분층이실현반궤,조정대마중적특정항적권치.실험결과표명,동시사용3충책략가이재제취벌치C상동적정황하,제고사전솔4%～28%,재사전솔불변적정황하,제고사준솔8%～30%.
In this paper,a retrieval method for traceability links between source code and Chinese docu-mentation is proposed,according to several features of Chinese software documents and source code.Based on the latent semantic indexing model,three strategies are used to improve:hierarchical en-hancement,identifiers classifying,and data dictionary.Experimental results show that all the three strategies could increase the recall by 4% to 28% under the same extracted threshold and the precision by 8% to 30% under the same recall.