Migrated from: https://wikimedia.mingle.thoughtworks.com/projects/language_engineering/cards/4306
Context
Apertium and potentially other MT engines are not designed well for keeping the html mark up of the input associated with correct words.
This issue was first explored by David in patchhttps://gerrit.wikimedia.org/r/#/c/143257/5
But it needs a broader scope and problem definition supported by enough test data to adapt it to other language pairs. This includes research and probably collaboration with Apertium since it is a generic MT problem
Narrative
As a <i>, I can </i>//<i>, so that </i>.//
Acceptance Criteria
//Enter the Acceptance Criteria here.//