Pretranlsation rules allow the system inserts TM matches and machine translation into the target text before the work even starts. This significantly speeds up the translation process and here are some points in which cases pretranslation can be helpful for you:
- When your document has a lot of translation memory matches. This will prevent you from having to click through and confirm each segment found in the TM.
- When your document has number-only segments. Smartcat will insert numbers and even convert their formatting according to language rules.
- For machine translation post-editing. Smartcat will fill every segment with the machine translation of the source text, for a translator to edit and confirm.
Let’s suppose you are going to pretranslate a project via both — a translation memory that has already been used to record translations and machine translation. By the way, the same pretranslation rules are applied to all existing and future documents within a project — so you don’t need to configure them for individual documents.
Open the Pretranslation tab.
If the source text contains a lot of numbers, click Source (Numbers Only) so that all the segments containing only numbers will be copied to the translation. Decimal and thousand separators, in this case, will be converted to match standards of the target language, at least in most cases.
You can specify whether the system must confirm such segments for every stage or only for the first one. Once segments are confirmed for a stage, project participants assigned will not be paid for editing these segments. And if the stage being confirmed is beyond the one that they are assigned to, the segments will be locked.
The rule for inserting numbers is always the first to execute.
Next, we want to use the translation memory. Press Add Rule.
Click Translation Memory.
Select which TM you want to use from the ones you have enabled on the project.
In the field Minimum match percentage, specify the minimum match percentage for a segment in the document and the translation memory at which the system must insert a translation from the TM.
You can also specify the minimum length of a segment (in words) at which the matches must be inserted using the Minimum word count in a segment field. This can be useful to eliminate the risk of inserting an improper wording.
Finally, specify in the Confirm segments field if the system should confirm the inserted translations and at what stage.
If you are sure of the quality of the translations in the applied TM and have set the minimum match percentage at 100%, you can specify that the system must confirm at the translation stage for such matches and have the editor to review such segments. If you have doubts about the quality of the TM units or have set the match threshold below 100%, you should not specify automatic confirmation of such matches since there might be segments required reviewing on the first stage. In this case, matches will be inserted, but the translator will check them, making corrections if necessary, and then manually confirming them. In cases where you have context matches (101% and above), you could even confirm segments for all the stages of your workflow.
You can also set up rules for each TM in a project (if there is more than one). The rules are executed one after the other; therefore, you should set the most reliable translation memory first.
By the way, more than one rule could be set up for each TM: for example, you can have Smartcat first insert 100% matches and confirm them, and then insert matches that are 75% and up, but without confirmation. Below, you may find an example with the following steps:
- Insert and confirm numbers and confirm them at the Editing stage.
- Insert and confirm 101% matches (context matches) longer than 2 words from your primary TM and confirm them at the Editing stage.
- Insert but don’t confirm 85% matches from your primary TM.
- Insert but don’t confirm 100% matches from another (say, domain-specific) TM.
To use the machine translation output to fill out the segments for which no matches have been found in the translation memories, click Add Rule and then choose Machine Translation.
Specify if you want the system to confirm segments containing machine translation. In most cases, these segments should not be automatically confirmed, because segments containing machine translation generally require corrections.
During pretranslation, the insertion of machine translation is the latest to process.
You can set up and run pretranslation later, but keep in mind that it will only be applied to those segments that have not yet been filled in.
After configuring pretranslation, press the Save & run button. Your project has been created.