Name Matching & Revision (Part 3)

Our second article about name matching and revision dealt with minimizing the risk of false negatives. This third article is about measuring the risk in practice. This is frequently done by testing two or three names, but this kind of layperson’s test has its pitfalls when it comes to compliance. Although this test is non-representative, it can still have relevance for compliance. A layperson’s test is most suitable for checking whether a collection takes into account the recent removal of sanctions. This is not a question of minimizing risk, however, but of checking the fundamental operating efficiency of a name matching system.

Both matches (positives) and non-matches (negatives) need to be taken into account for a risk assessment. We must also differentiate between correct matches (true positives) and incorrect matches (false positives), as well as between correct non-matches (true negatives) and incorrect non-matches (false negatives). We want as few false negatives as possible, as these mean that potential risks are not being recognized. However, a high number of false positives is also undesirable if a lack of resources means that these cannot be verified and remain open issues.

Eurospider possesses a comprehensive list of pairs of different names that refer to the same person. For example, “Boris Yeltsin” and “Boris Nikolayevich Eltsin” both refer to the former Russian president. There is a significant difference between these two names, creating a challenge for any name matching system.