Reconstructions of systems often provide useful insights, and the work here is no exception. A description is first given of how the authors reconstructed NASA’s automated requirements measurement (ARM) tool. Then, as the authors discuss and compare tool output, several useful insights emerge. For example, the ARM tool is suspected to have lost information embedded in tables. Also, the ARM tool was determined to be flawed in the way it recognized subjects and paragraphs. These flaws, however, may not have manifested in practice at NASA, where document production was likely to have been highly prescribed.

Figures 5 to 8 show that in terms of counts of quality indicators, the reconstruction closely matches the original tool’s output, providing evidence that the reconstruction was indeed successful. In contrast, figures 9 to 12 reveal considerable differences in numbering structure counts, but this is explained by the fact that the original tool required an actual number to precede a requirement statement, whereas the reconstruction does not. Counts by depth in figures 9 to 12 are described as more accurate for the reconstruction due to the way it tracks depth. Section 4 discusses various suggestions for improvement, including the idea of integrating requirements critiquing directly into requirements writing.

The generated quality indicator data is not analyzed for the rate of false positives, and arguably the most useful insight of all is missing. However, the authors have made their reconstruction available online (http://www.kuwatche.com/ARM.html) for others to make such determinations. This paper, therefore, is strongly recommended to those working in requirements engineering.