A fuzzy match is a translation memory segment that closely resembles new source text without being identical, scored as a percentage of similarity.
Why Fuzzy Match Matters
Fuzzy matching matters because most content is never repeated word-for-word — a product description gets rephrased, a UI string changes one word, a release note swaps a feature name. Exact matches alone would miss almost all of this near-duplicate text, leaving translation memory far less useful. Fuzzy matching closes that gap by comparing new source text against every segment already stored in TM and returning a similarity score, so close variants still surface as suggestions instead of being retranslated from scratch.
The percentage threshold controls what a linguist or AI sees. Segments similar to a stored TM entry above roughly 75% are surfaced automatically during editing; a linguist can also search TM manually at any threshold to find weaker matches. Fuzzy match should not be confused with a 100% basic match — a 100% basic match is identical text in a different context and still needs a quick review, while a fuzzy match is genuinely different text that needs actual editing to fit.
For professional translation services, fuzzy matches carry a scaled discount: segments between roughly 50% and 99% similarity are billed at a reduced per-word rate that increases as similarity increases, while anything below 50% is charged as new translation. For AI translation, by contrast, fuzzy matches are not a pricing lever — the AI uses them as reference to stay consistent with prior wording, but the bill is always the subscription’s word-count tier, not the match percentage.
A Concrete Example
A source segment “Download the app” is already stored in translation memory. A new document contains the sentence “Download our app” — different by one word, so it scores roughly 92% similar. Instead of treating this as brand-new text, TM surfaces the stored translation with the difference highlighted, so a linguist only needs to adjust “the” to “our” rather than translating the sentence from nothing. Under professional services, that segment is billed at a discounted fuzzy-match rate instead of the full new-word rate.
How Fuzzy Match Works in Taia
In Taia, fuzzy matches at 75% or higher are shown automatically inline while translating in the CAT editor, and linguists can manually search translation memory for matches below that threshold at any time. Fuzzy matching is part of Taia’s translation memory system, available on the Pro plan, and applies the same scaled-discount logic to every professional services project.