The ambiguity finding tool SREE uses only lexical analysis to ensure 100 percent recall. SREE operates by searching for the presence of words and phrases defined in ten subcorpora. For example, the quantifier subcorpus contains words like few, many, several, and some as indicators of ambiguity.

To allow others to learn from their mistakes, the authors make clear the weaknesses of SREE. The precision of SREE’s analysis of two requirement specifications was found to be 66 percent and 68 percent respectively. A mixed SREE/manual inspection approach was estimated to take about 75 percent of the time that would be needed for a purely manual inspection--not a large saving of time. In Section 6.3, the reader’s attention is drawn to the difficulty of dealing with plural nouns, and suggestions are made for modifications to SREE.

Some researchers would question the method used to estimate the time needed for the mixed SREE/manual inspection approach. If a requirements specification has to be read over for ambiguities that SREE does not address, one could argue there is actually no saving in time. My main criticism is that the paper lacks a breakdown of precision by the ten subcorpora. The first author’s thesis [1] makes clear that the plural subcorpus is responsible for some three-quarters of the false positives. Excluding this subcorpus would greatly improve the tool’s precision and the likelihood of such a tool being adopted in an industrial context. Despite these criticisms, this paper is strongly recommended to those working in requirements engineering.