Development of a document traceability analysis methodology for SIS-RT

In this study, we propose a document traceability analysis methodology for high-reliability software inspection tool, SIS-RT. The proposed methodology calculates the similarity between two sentences using the cosine vector similarity formula. The keywords are derived based on inverse document frequencies. When we apply the proposed methodology to real documents, it is found that a source sentence and a destination sentence can be determined to have traceability when the similarity between the two sentences is over 80%. In case the similarity between two sentences is over 30% but less 80%, it turns out that the decision of the inspector is required. Even though sometimes the trends between the similarity of two sentences and the traceability of two sentences do not exactly matches, it is found that the methodology we propose in this paper is useful for traceability analysis