Every Translation Memory (TM) has its own search tool for finding stored segments quickly — whether you're checking if a sentence has already been translated, cleaning up duplicate entries, or verifying how a specific term was rendered before.
Open the Translation Memory you want to search (see All about Translation Memories if you need help finding or creating one) — the search tools live directly on its Overview page, under the Search card.
​

💡 Who is this for? This guide is for Account Admins and anyone else granted permission to manage Translation Memories, who need to search a Translation Memory's stored segments within the wxrks platform.

wxrks gives you three ways to search a Translation Memory, each suited to a different kind of lookup:

Full Text Search also adds a language dropdown so you can scope the search to the source language or a specific target language, plus the Rows selector for how many results show per page.
​Fuzzy Search and Contains Search always search from the TM's source language and load every match at once — there's no language dropdown or Rows setting for those two modes.

Full Text Search is the default mode and the one you'll use most often. Type a word or phrase and wxrks tokenizes it and applies stemming — reducing words to their root form so different forms of the same word count as a match. Searching translate, for example, also surfaces segments containing translation or translators, without you having to type every variant.

Fuzzy Search compares your query against the full text of every stored segment and returns a percentage match, instead of a simple contains-or-not result. Reach for it when you want to know how closely a new sentence resembles something already in the TM — it's the same kind of comparison the wxrks Editor uses to surface translation suggestions.
Drag the Threshold slider to set the cutoff: only segments scoring at or above it are returned. The lowest selectable threshold is 49%; searches start at a default of 75%.

Searching for a segment's exact stored text returns a 100% match:

Searching for only part of that same segment — leaving off the last word — dropped the match to 67.65%. At the default 75% threshold, that result wouldn't have appeared at all; the screenshot below only shows it because the threshold was lowered first. If a fuzzy search comes back empty, a low threshold is often why — try dragging it down before assuming the segment isn't there.

Contains Search looks for the literal text you type, with no tokenizing or stemming — useful when Full Text Search's word-matching is too broad, or when you need to find text containing punctuation, symbols, or a specific pattern.

Turn on the .* toggle — only available in Contains Search — to search with a regular expression instead of plain text. It's the fastest way to find segments matching a pattern, such as .*TMS.* to match TMS wherever it appears in a segment.

Click the funnel icon to the left of the search box to filter by more than just text. The panel lets you narrow results by when a segment was created or last modified, the filename it was imported from, or who created or last modified it — useful for tracking down segments from a specific import or teammate without scrolling through the whole TM. Applied filters show up as small chips below the search box (like the Text chip in the screenshots above).

Columns controls which target languages appear in the results table — useful when a TM has several target languages and you only need to check one or two. Click Columns, then check or uncheck each language; Select All / Deselect All toggles them all at once.

Each result row shows the source segment alongside its translation in every visible target column. A few details worth knowing:

Related articles: All about Translation Memories, How to add new segments to Translation Memories, How to create a Translation Memory.