Pre-translation is an important part of the project creation process. If done properly, this process will save the translator time and, accordingly, will save the customer money on translation projects. In addition, it serves to enforce consistency as segments can be filled with approved translations and in some cases locked so that translators and/or editors cannot modify the text. The ideal process will give the project managers many options on how they want to process files and automate this step as much as possible.
The process will vary depending on the CAT tools but for this article we will concentrate on comparing smartCAT with memoQ, as it is one of the leaders in the industry.
Here is a summary of the differences between the way smartCAT and memoQ handle the pre-translation process.
- The pre-translation process is part of the project creation in smartCAT. It is the last step of the 3-step Wizard. With memoQ, the project manager will first need to create a project and then apply the pre-translation rules. Although not a major difference, it does impose an extra step in the file preparation process. smartCAT offers the flexibility to let the project manager set up pre-translation rules during project creation but also after the project has been created. This will for example allow for running the process again in the case that a new TM has been added to the project.
- With memoQ, it seems to be only possible to choose one type of match for the pre-translation:
For example, you can pre-translate the file using context matches, 100% matches or what memoQ calls “Good matches” (95% and higher — that threshold can be modified) but you have to choose only one of the options. Once you choose an option, all the matches above that threshold will be pre-translated automatically. By checking the “Use Machine Translation” checkbox, all the empty segments will be translated using the MT engine which is designated as “Preferred” in the memoQ settings. These segments will not be confirmed and get a special status. Another disadvantage is that it is not possible to set a priority list for the TM during that stage. The matches will be picked from the TM in the order that they are listed in the project settings. In contrast, with smartCAT you can perform as many separate lookups as you need: context matches, 100% matches and anything down to 75%, which gives you more confirmation options. And you can prioritize the TMs accordingly to your needs during the process.
Here’s an example of a pre-translation setup in smartCAT:
In this case, two different TMs are selected for the pre-translation process and the project manager is free to select the order in which the matches will be applied, even if the order varies from what was selected in the previous step. Also note that with smartCAT you can select to confirm segments at different stages of the translation process. In the screenshot above, exact matches (102%) are confirmed at the Proofreading stage, while the 100% matches from a different TM are inserted but not confirmed. Clearly, this is not a typical setup, but it serves as an example of the flexibility of the smartCAT options.
memoQ does allow you to select a scope for pre-translation: All files in the project, selected files, currently opened document, a selection of segments, all segments under the current segments and also views (“view” is a meta-document in memoQ which contains segments from one or multiple files which have a set of criteria, such as segment status, tags, text, change date, repetition, comments, etc). The pre-translation rules you can choose in smartCAT can only be applied to all the files in the project. So, in that regard memoQ is a bit more flexible.
- On the other hand, the locking mechanism seems to lack flexibility:
Once again, you can lock only one type of matches and change the status of these matches only. You don’t seem to have the option to lock different types of matches for different stages of the process, as can be done in smartCAT, or change their status. As shown above, smartCAT allows the project manager to choose what type of matches will be locked for each step of the translation process. In another example, 99% matches can be pre-translated and marked as translated so that they will appear locked for the translators. Such a choice would greatly reduce translation costs without compromising the quality. However, this is not possible in memoQ. Based on the project requirements, in smartCAT, other types of matches can be locked for the editors and/or marked done for the proofreaders. The smartCAT implementation is clearly more flexible and gives the project managers more options.
- Where memoQ offers more options for now is that it allows the PM to create versions of files. For example, create a snapshot of the files before pre-translation. This could be useful if pre-translation doesn’t match expectations. Also, the word count analysis can be made to ignore segments that have been locked. This feature has been recently added to smartCAT as well.
- There are some features in memoQ that are also clearly more advanced than in smartCAT. One such pre-translation feature is the ability to use a TM-driven segmentation. For example, you have the two following segments in your file:
Seg #1: I have an apple.
Seg #2: The apple is delicious.
If in the TM you have the “I have an apple. The apple is delicious.” match from a similar project with different segmentation rules, memoQ will join the segments to insert this match, if TM-driven segmentation is enabled. It is also true if a segment needs to be split.
6. Fragment assembling is another good feature of the memoQ pre-translation. This is how it works. If in your file you have the following segment:
I have an apple.
you might have dozens of segments like “I have a pear”, “I have a car”, “I have a bottle of delicious Russian vodka”, and so on in your TM. If you also have a segment “Apple” in your TM as a separate word, memoQ tries to assemble the translations based on the statistical splitting of the segments into phrases. In French, the assembled translation might look like:
In this case, the French translation still needs some editing, but the framework of the sentence is inserted which does save the translator some time when the results are as good as this one.
The pre-translation processes in these two applications differ in their implementations. smartCAT affords projects managers more flexibility on how to pre-translate full segments and lock pre-translated segments automatically, given different criteria. Being able to prioritize TMs and choose different pre-translation and locking options should generally result in less work for the translation teams. Some CAT tool users will sometimes pre-process files manually in order to lock segments and save money on translation. smartCAT can automate this process completely. The pre-translation process is also an integral part of the project creation, as opposed to a separate process in memoQ.
Where memoQ is stronger is when it tries to create new matches by assembling portions of existing segments as described in the example above. This could in some cases generate additional matches but these matches will need to be reviewed carefully by the translation team. Project managers might favor the smartCAT process for the additional options it offers that can match complex localization processes. Translators, on the other hand, might appreciate the segment assembling features of memoQ.