Enhanced abbreviation-expansion pair detection for glossary term extraction

Context: Providing precise definitions of all project specific terms is a crucial task in requirements engineering. In order to support the glossary building process, many previous tools rely on the assumption that the requirements set has a certain level of quality. Yet, the parallel detection and correction of quality weaknesses in the context of glossary terms is beneficial to requirements definition. Objective: In this paper, we focus on detection of uncontrolled usage of abbreviations by identification of abbreviation–expansion pair (AEP) candidates. Methods: We compare our feature-based approach (ILLOD+) to other similarity measures to detect AEPs and propose how to extend the glossary term extraction (GTE) and synonym clustering with AEP-specific methods. Results: It shows that feature-based methods are more accurate for AEPs than syntactic and semantic similarity measures. Experiments with PURE data-sets extended with uncontrolled abbreviations show that ILLOD+ is able to extract abbreviations as well as match their expansions viably in a real-world setting and is well suited to augment previous synonym clusters with clusters that combine AEP candidates. AEP clusters generated with ILLOD+ are generally smaller than those based on syntactic or semantic similarity measures and have a higher recall. Conclusion: In this paper, we present ILLOD+, an extended feature-based approach to AEP detection and propose a workflow for its integration to clustering of glossary term candidates to enhance term consolidation in evolving requirements.