This document proposes a method called "Reversed analysis" to measure machine translation (MT) post-editing effort. The method involves applying a traditional translation memory (TM) leverage analysis technique to the target content of a raw MT output compared to a virtual TM of final post-edited content. This allows the method to evaluate each MT segment and calculate a cost reduction for translation jobs based on the matching percentage to the post-edited TM. The document describes applying this method using the ETMA (SDL TMS 2011) tool.
FR = Full RateSame principle for both TM leverage and MT: By generating/proposing a “pre-translation” to the Translation supplier (= “raw MT” in the case of MT technology; = TM leveraged segment in the case of TM technology), the effort spent to get to the final translation/translated content is reduced.In ETMA, both technologies/processes are combined: TM leverage on fuzzy matches segments, MT on ‘ New words’ = ‘ No match’ segments.
aka Post-editing analysis.*: An important aspect is to use the same TM matching algorithm as the one that is used for evaluating and pricing the TM matches. Since the algorithm is there you do not need to buy a new tool. And it is well understood from all parties involved (since it used for calculating the translation costs).*: As a first pass, the cost model used could be the same as the one used for TM leverage analysis. A refined “reverse analysis” cost model can be obtained wit h the help of Productivity gains evaluation (see slide xx) .
(*): Alternate method is to use the existing bilingual files (if there is not a mix of MT and TM segments and if the ITD before any translation work can be downloaded pro-actively).
(*): Alternate method is to use the existing bilingual files (if there is not a mix of MT and TM segments and if the ITD before any translation work can be downloaded pro-actively).TODO: ADD A SECTION describing the PEMT Prod Gain Evaluation (see the slide deck presented to Suppliers)