I maintain a concept-oriented EN>KO termbase and normalize entries to lemmas with inflected variants and forbidden terms, but TBX exports between memoQ 10 and Trados 2022 keep breaking case sensitivity and locale tags. Any dependable resources or workflows — regex patterns, lemmatization rules, or QA passes in Xbench/Verifika — to reduce false positives on pairs like log-in/login/sign-in (I logged 37 last week) and handle US vs UK variants without over-flagging?