2024-03-29T22:54:08Zhttps://ipsj.ixsq.nii.ac.jp/ej/?action=repository_oaipmhoai:ipsj.ixsq.nii.ac.jp:002126972023-11-14T00:51:14Z06164:06165:06522:10650
Building SOFL-to-Java Traceability Links using Multi-dimensional Similarity MeasuresBuilding SOFL-to-Java Traceability Links using Multi-dimensional Similarity Measureseng形式手法http://id.nii.ac.jp/1001/00212591/Conference Paperhttps://ipsj.ixsq.nii.ac.jp/ej/?action=repository_action_common_download&item_id=212697&item_no=1&attribute_id=1&file_no=1Copyright (c) 2021 by the Information Processing Society of JapanHiroshima UniversityHiroshima UniversityHosei UniversityJiandong, LiShaoying, LiuRunhe, HuangTo achieve an automatic formal specification-based program fault detection, the open problem of how to automatically link the components in the formal specification to the corresponding ones in the implemented code must be addressed. To reduce the manpower and time cost, some automated techniques have already been developed but their effectiveness is limited mainly due to the over dependency of textual similarity. In this paper, we present an automatic method for constructing the traceability links between SOFL formal specifications and Java program code. Unlike the existing work, our method not only considers the semantic similarity, but also structural, functional, and relational similarities as the measurement dimensions. It also operates at multiple levels of a specification, such as data flows, processes, and modules, to establish fine-grained link relationships between artifacts. Further, we conduct a comprehensive empirical evaluation of the proposed method using selected two modules of a critical ATM system's SOFL formal specification and its Java implementation with the size of 951 code of lines and 36 traceability links. The result shows that we can establish SOFL-to-Java links with the precision of 97.2% which is close to highest accuracy of existing naming convention technique in the situation of consistent identifier and the precision of 88.8% illustrating high performance in precision and generality in the situation of inconsistent identifier.To achieve an automatic formal specification-based program fault detection, the open problem of how to automatically link the components in the formal specification to the corresponding ones in the implemented code must be addressed. To reduce the manpower and time cost, some automated techniques have already been developed but their effectiveness is limited mainly due to the over dependency of textual similarity. In this paper, we present an automatic method for constructing the traceability links between SOFL formal specifications and Java program code. Unlike the existing work, our method not only considers the semantic similarity, but also structural, functional, and relational similarities as the measurement dimensions. It also operates at multiple levels of a specification, such as data flows, processes, and modules, to establish fine-grained link relationships between artifacts. Further, we conduct a comprehensive empirical evaluation of the proposed method using selected two modules of a critical ATM system's SOFL formal specification and its Java implementation with the size of 951 code of lines and 36 traceability links. The result shows that we can establish SOFL-to-Java links with the precision of 97.2% which is close to highest accuracy of existing naming convention technique in the situation of consistent identifier and the precision of 88.8% illustrating high performance in precision and generality in the situation of inconsistent identifier.ソフトウェアエンジニアリングシンポジウム2021論文集20211431502021-08-302021-08-26