Improving Dynamic Trace Retrieval through the Use of Phrasing

Dynamic traces retrieval utilizes information retrieval techniques to automatically generate links as opposed to manually constructing and maintaining links using traditional traceability methods. However, the dynamic method is facing a challenge of low precision in the trace retrieval results. This paper describes a strategy to improve the precision of the retrieval results by integrating phrases detected from requirements using a part-of-speech tagger. Results on various datasets illustrate this strategy significantly improves the accuracy of the retrieval by generally increasing the overall precision, and meanwhile pushing more good trace links to the top of the retrieved links list.