Toward Improved Artificial Intelligence in Requirements Engineering: Metadata for Tracing Datasets

Abstract

Data is the driver of artificial intelligence in requirements engineering. While some applications may lend themselves to training sets that are easily accessible (such as sentiment detection, feature request classification, requirements prioritization), other tasks face data challenges. Tracing and domain model building are examples of applications where data is not easily found or in the proper format or with the necessary metadata to support deep learning, machine learning, or other artificial intelligence techniques. This paper surveys datasets available from sources such as the Center of Excellence for Software and Systems Traceability and provides valuable metadata that can be used by re-searchers or practitioners when deciding what datasets to use, what aspects of datasets to use, what features to use in deep learning, and more.