IEEE Transactions on Software Engineering 47 (6): 1164-1183 (2021)
Metamorphic Robustness Testing:
Exposing Hidden Defects in Citation Statistics and Journal Impact Factors

Zhi Quan Zhou 2 , T.H. Tse 3 , and Matt Witheridge 4
[free download from IEEE Xplore | free download from IEEE digital library |
ICSE '20 journal-first presentation in YouTube | press release |
reported in ScienceAlert, Nature Index, Communications of the ACM, Psychology Today, The Australian, and Wikipedia]

data-hide-no-mentions="true" data-link-target="_blank" />

We propose a robustness testing approach for software systems that process large amounts of data. Our method uses metamorphic relations to check software output for erroneous input in the absence of a tangible test oracle. We use this technique to test two major citation database systems: Scopus and the Web of Science. We report a surprising finding that the inclusion of hyphens in paper titles impedes citation counts, and that this is a result of the lack of robustness of the citation database systems in handling hyphenated paper titles. Our results are valid for the entire literature as well as for individual fields such as chemistry. We further find a strong and significant negative correlation between the journal impact factor (JIF) of IEEE Transactions on Software Engineering (TSE) and the percentage of hyphenated paper titles published in TSE. Similar results are found for ACM Transactions on Software Engineering and Methodology. A software engineering field-wide study reveals that the higher JIF-ranked journals are publishing a lower percentage of papers with hyphenated titles. Our results challenge the common belief that citation counts and JIFs are reliable measures of the impact of papers and journals, as they can be distorted simply by the presence of hyphens in paper titles.

Keywords: Metamorphic robustness testing, metamorphic testing, fault-based testing, software robustness, oracle problem, citation count, journal impact factor, Scopus, Web of Science, Google Scholar, verification and validation.

1. This work was supported in part by a linkage grant of the Australian Research Council (Project ID: LP160101691) and an Australian Government Research Training Program scholarship. We wish to thank Morphick Pty Ltd for supporting this research. We are grateful to the anonymous reviewers for their valuable comments.
2. (Corresponding author.)
School of Computer Science and Software Engineering, University of Wollongong, Wollongong, NSW 2522, Australia.
3. Department of Computer Science, The University of Hong Kong, Pokfulam, Hong Kong.
4. School of Computer Science and Software Engineering, University of Wollongong, Wollongong, NSW 2522, Australia.


  Cumulative visitor count