Recommending Terms for Glossaries: A Computer-Based Approach

Abstract

Glossaries in Software Requirements Specifications (SRS) aim at establishing a common ground of definitions. However, ambiguous terms as due to tacit knowledge are seldom captured in glossaries. In addition, even if they are captured, they are seldom read, because potential readers are convinced that they already know how the term is defined. Such misunderstandings introduce high risks in projects - especially because they are so hard to detect. Therefore, a trigger is needed to start a discussion about these potentially dangerous terms. In this paper we show how context aware requirements engineering tools can heuristically detect these terms and point out the risk attached. We introduce two simple, yet powerful heuristics: Occurence counting detects important terms, comparison with old glossaries detects terms that others found worth defining in a glossary. Thus, we make use of glossaries from past projects to suggest possible terms of interest for current projects. Our approach was implemented and applied to six software projects. Based on these experiences we show the effectivity of our heuristics and how they could be used by learning organizations to reduce such ambiguity risks in their specific domain.