4 Narrative Fuzzy Matching
Narrative Fuzzy Matching refers to applying fuzzy matching logic in narrative fields for SWIFT MT and ISO 20022 payment messages which has only been set to exact previously. Narrative tags are free-text fields containing information provided by the sender.
These contain valuable information, which is required for screening. However, because of spelling errors, abbreviations, or transliteration differences, exact keyword matching fails to capture risk-sensitive information. The application of fuzzy logic against the narrative helps identify close variants of names, terms, or sanctioned entities.
- Using GenAI-based keyword extraction
- Without using GenAI-based keyword extraction
- Combining both GenAI-based and non-GenAI keyword extraction methods
- When fuzzy logic is applied in narrative fields, the number of matches is expected to increase. Therefore, it is advised to perform a calibration cycle for this use case during implementations.
Table 4-1 Indicative Results after enabling Fuzzy Matching for Narrative Fields
| Example | The Narrative Text | Number of Matches using Exact Match logic | The number of matches using Fuzzy Matching at 60 Threshold | The number of matches using Fuzzy Matching at 85 Threshold |
|---|---|---|---|---|
| 1 | XY90/ Wire for BIN HANI Laseed | 21 matches - True match missed | 1024 matches | 134 matches |
| 2 | ALL NOTICES, CERTIFICATES OR OTHER COMMUNICATIONS TO THE APPLICANT WILL BE DELIVERED TO: AL LINURES Angle, NY | 32 matches - True match missed | 1405 matches | 70 matches |