Answer: About as well as an educated person who speaks the supported language, but has no specific knowledge of the unsupported language would do matching the names.
Example: RNI supports English but not Indonesian (which is written using the same alphabet as English).
For use-cases that occasionally include Indonesian-origin names (e.g., US Border Control, International Financial Watchlist-Monitoring), telling RNI that those names are "in English" can be done while retaining world-leading accuracy in the aggregate. This is because:
- RNI's culture-agnostic algorithms are robust for names that cross cultures (e.g., a German given name with a Chinese family name) as compared to culture-specific algorithms that can be more brittle, and
- the names for these use-cases often have received some English-oriented normalization (e.g., titles like "Mr." rather than "Pak" and inclusion of both a given and family name)
For use-cases that include predominantly Indonesian-origin names RNI is likely insufficient because in those use-cases more origin-specific matching phenomena are likely to be displayed, whether reflecting local pronunciation, nicknames, or omission of certain name words.
We aspire to add more culture-specific knowledge to RNI while maintaining its robustness to cross-culture names and would be glad to hear of specific example use-cases to inform this.