Translation Memory, Matching, Alignment, and Data Exchange
Andrew Rothwell, Joss Moorkens, María Fernández-Parra, Joanna Drugan, Frank Austermuehl · 2023
Almost all CAT tools store legacy translations in TM databases, while some of the most popular tools offer retrieval from both TM and bitexts. Detail of Phrase&s;s match proposal pane showing three English to Chinese matches. In a box underneath them the differences between the active SL segment and the SL segment of the top proposal are marked as tracked changes. CAT tools are often said to be language-independent , i.e. they work in the same way and equally well with whichever languages they are used. The equivalence principle that underlines all leveraging of legacy translations is implemented in CAT tools through the mechanism of statistical alignment, first formulated by Gale and Church as an algorithmic method for constructing parallel corpora for use in Natural Language Processing applications. Sophisticated searching and matching algorithms have been developed to perform the edit-distance calculations that allow the extraction of re-usable translation units from CAT corpora.