Customising fuzzy thresholds
Managers can adjust the threshold for fuzzy scores in settings. This threshold will be applied to the entire team. Strise will then present all PEP and Sanctions results that have at least the given fuzzy score against the name in Strise.
Managers can also choose to eliminate false positives by filtering out hits where metadata (such as date of birth, nationality, etc.) does not match.
Sanctions: Searching & matching
Strise retrieves sanction data from the following sanction lists in XML format:
We then consolidate these lists into our sanctions dataset and search for matches. We use the Damerau-Levenshtein distance algorithm to perform fuzzy scoring, with the following features:
- Upper and lowercase letters are not significant. "StRiSe" = "strise".
- The order of words is not significant. "Erna Solberg" is the same as "Solberg Erna". "Erna Solberg" has a score of 0.92 against "Stolberg Erna" and "Erna Stolberg".
- All accents and special characters are removed. "Stèin-Erík" becomes "Stein Erik".
- Matching on subsets of names is supported. "Vladimir Vladimirovich Putin" is the same as "Vladimir Putin".
- Matching on subsets of names have some limitations to avoid unnecessary false positives. The subset of two names are only matched if: both names have 3 or more words, or both names have similar lengths, or one name has 2 words and the other 3.
- All aliases used by Strise will be screened against all aliases of sanctioned entities and PEPs.
- All IDs used by Strise will be screened against PEP and sanctions lists. However, a perfect match is required, regardless of the fuzzy threshold set in the settings. Accents and special characters are removed from the IDs before searching. "123-456-abc" is the same as "123456abc".
- Date of birth (d.o.b) can be used to prevent false positives from name matching alone. Sven Anders Olsson (Feb 8, 1943) is sanctioned, while Sven Anders Olsson (May 26, 1966) is not.
- If the entity being screened has a full date of birth (year and date) in both Strise and the sanction list, we will use that full date to filter out false positives.
- If the sanctioned entity only has a birth year, then we will only use the birth year from the entity we're screening to filter out false positives.
- If the sanctioned entity only has an interval of birth years, then we verify that the birth year of the screened entity is within that range and use this to filter out false positives.
All the mentioned rules are combined to perform a search. For example:
- "Vlàdimír-Vladimirovich Putin" is equivalent to "Putin Vladimir"