Alignment is a procedure to generate translation memories from pairs source documents and their translations, called target documents.
Alignment usually has three steps:
- A software tool, usually referred to as an aligner, automatically segments the source and target documents, aligns (pairs) the source and target segments and displays the result, typically in a table.
- A linguist checks the result and corrects misalignments.
- Once all segments are properly aligned, the software aligner generates a translation memory containing all the aligned segment pairs as translation units (TUs).
The reliability of automatic software segmentation and alignment depends on both the aligner and the level of similarity in linguistic structures and document formatting between the source and target documents. With larger documents, the human part of alignment can be tedious and time-consuming, unless the result of automatic alignment is near perfect.