When memoQWeb (legacy) pre-translates a document, it looks up every segment in the translation memories and LiveDocs corpora in the project, and inserts the best match. If there is no match, memoQWeb (legacy) can put the segment together from fragments, or machine-translate the source segment. It can also try to join or split segments automatically, if that results in better matches.
You can control the quality of the matches you use.
In addition, memoQWeb (legacy) can save the version of a document before pre-translation starts, so you can check what happened during pre-translation, and return to the previous version if necessary.
Caution: You may not be able to pre-translate a document in an online project while memoQ server is running an automated task on the same project. In this case, you will receive a message telling you to wait a few minutes. This is not an error: you can continue with pre-translation a little later.
How to get here
- Sign in to memoQWeb as an administrator or a project manager.
- On the Active Projects list, click a project.
- At the top of the screen, click the Translations tab.
- In the list, select one or more documents.
The Pre-translate page opens.
What can you do?
Learn details about the documents
The info bar at the top shows:
- The number of selected documents. To see the documents' names, hover over the blue information icon.
- The project's source and target languages
- The number of segments in the documents
- The number of words (raw and weighted) in the documents.
- The number of characters (raw and weighted) in the documents.
Choose what matches memoQWeb should insert
Use the dropdown in the Lookup tab, under the TMs and LiveDocs corpora section.
- Exact TM or corpus match with context: Pre-translate a segment only if there is a context (101%) or a double-context (102%) match.
- Exact TM or corpus match: Pre-translate a segment only if there is an exact (100%) match or a context match.
- Good TM or corpus match: Pre-translate a segment only if there is a match above the "good match threshold". This is defined in the TM settings. To check this setting:
In the memoQ desktop app, open the project. If the autopilot is on, turn it off. On the left, click Settings.Click the TM settings icon, and edit the TM settings profile that is active in the project.
- Any TM or corpus match: Always pre-translate a segment if there is a match, no matter the match rate. In fact, it must still be above minimum match threshold (60% by default). To use this, you must turn off TM-driven segmentation (see below).
Choose the translation memories and LiveDocs corpora to use for pre-translating
There are more options to fine-tune the process:
Only unambiguous matches: If you check this check box, memoQWeb will pre-translate a segment only if there is exactly one exact (100%), context (101%), or double-context (102%) match. If there are two or more, the segment will not be pre-translated. You must choose Exact TM or corpus match with context or Exact TM or corpus match
from the dropdown.
Perform fragment assembly: If there is no match from the translation memories or from the LiveDocs corpora, memoQWeb can insert a fragment-assembled match. To make it work,
turn on this switch. To fine-tune the details, use the check boxes below.You can choose what counts as a fragment, and how memoQ should put together a fragment-assembled match from them.
Under Matches to include, you can choose to replace terms, non-translatable items, numbers, results of auto-translation rules, and fragments from translation memories and LiveDocs corpora. Normally, memoQWeb will use all of these to replace parts of the source text. If you do not want to use one or another, clear the check box of one of these items.
Next, choose how much the fragment-assembled match should cover from the source text. Choose one of the radio buttons under Suggest only:
- Full matches covered by one single hit: The fragment-assembled match must cover the entire source segment, consisting of one single fragment match. Practically, this turns fragment matching off.
- Full matches covered by several hits: The fragment-assembled match must cover the entire source segment. This means that every word of the source segment must be replaced by something from the resources.
- Matches with a coverage of at least: Only part of the source text must be replaced. To set how much, choose a percentage in the box next to the radio button. Normally, memoQ offers fragment matches if it can replace more than 50% of the source text.
You can have fragment-assembled suggestions that does not contain any source text. In this case, text that cannot be replaced will not be there in the suggestion. For this, check the Delete source text without any match check box.
Normally, memoQ adjusts the lowercase/uppercase setting of the terms to the source text that is replaced. To stop this, check the Do not change the case of terms check box. When this is checked, memoQ will keep the lowercase/uppercase setting from the term base.
Use machine translation if there's no TM match: If there is no match from the translation memories or from the LiveDocs corpora, memoQWeb (legacy) can insert a machine translation of the source segment. To do that, it will use the machine translation services set up in the project's MT settings resource. To make it work,
turn this switch on. To set up the MT settings resource, use the Resource Console's MT settings page in the memoQ desktop app. Then add the resource to the project on the Settings page in memoQ online project.
When you turn the switch on, settings for each target language appear.
If the check box for a language is cleared, memoQWeb does not do machine translation for that language. If a check box is checked, its dropdown shows the MT services set up in the project's MT settings resource. Choose one, and memoQWeb will use that MT service for pre-translation to that language.
Machine translation does not work if TM-driven segmentation or fragment assembly is turned on.
Use TM-driven segmentation: memoQ can automatically join or split segments if there are better matches that way. If you are pre-translating for good (or better) matches only, memoQWeb (legacy) can try to join or split segments to see if there is an exact match that way. To do this,
turn on the Automatically join and split segments for best match switch. Then, if there is no exact match for a segment, memoQWeb (legacy) will check if the segment plus the next ones offer a good or an exact match. If joining does not work, but there is an exact match for the first part of the segment, memoQWeb (legacy) will split the segment after this exact match, and it will continue to examine the second part of the split segment for automatic joining and splitting. To fine-tune the details, use the settings below.
- Prevent segments from fragmentation: Normally, memoQWeb (legacy) does not split one or two words from a segment. To set the number of words memoQWeb (legacy) can split: Change the Minimum segment length after splitting (words) setting.
- Allow memoQ to join more than two segments: To get a better match, memoQWeb (legacy) will join two segments and stop there. If the document is too fragmented, you can allow memoQWeb (legacy) to join more segments, if it can get a much better match that way. To allow this, set the Maximum number of segments to join setting higher than 2.
- If memoQWeb (legacy) is allowed to join more than 2 segments: Normally, it will join one more segment if that makes the match at least 10 percent better. To change this improvement condition, change the Minimum match improvement (%) setting.
TM-driven segmentation does not work if you have machine translation or fragment assembly turned on.
You can tell memoQWeb (legacy) which
TMs and corpora: Click the Select TMs and LiveDocs corpora link. The Select TMs and LiveDocs corpora for pre-translation window opens:
memoQWeb (legacy) uses the TMs and corpora which have their check boxes checked. Check or clear the check boxes for each TM or corpus as needed. To select or deselect all the TMs or corpora, click the Select all or Deselect all links.
To save your changes and return to the
To return to the
Confirm and lock segments that were successfully pre-translated
memoQWeb (legacy) can automatically confirm or lock segments that were pre-translated. Only confirmed or proofread segments count in progress information. A document is 100% complete only if all segments are confirmed or proofread. If you do not want others to change the translation, you can also lock these rows.
Do this in the Confirm/Lock section on the page.
First, tell memoQWeb (legacy) which segments should be confirmed or locked. Use the radio buttons under Rows to confirm/lock based on match rate:
- Do not confirm or lock segments: Does not confirm pre-translated segments. They remain in the pre-translated status.
- Exact match with double context: Confirms and/or locks double-context (102%) matches.
- Exact match with context: Confirms and/or locks context (101%) or better matches.
- Exact match: Confirms and/or locks exact (100%) or better matches. Be careful with this setting. Without the context, pre-translation may insert inadequate translations.
- Only if match is unambiguous: If this is checked, memoQWeb (legacy) will not confirm or lock a segment if there are two or more exact or better matches for it.
Second, tell memoQWeb (legacy) what should happen to a segment that has the kind of match you are looking for. Use the
- Pre-translated: Does not change the status of the segments.
- Translator confirmed: Confirms the segments as Translator confirmed.
- Reviewer 1 confirmed: Confirms the segments as Reviewer 1 confirmed.
- Reviewer 2 confirmed: Confirms the segments as Reviewer 2 confirmed.
- Make rows locked: If you check this check box, memoQWeb (legacy) will also lock the segments it confirms.
Save the documents' contents before or after pre-translation: Create a new version
memoQWeb can take snapshots of a document before and after pre-translation. This way you can check the changes, or return to the previous version if anything goes wrong.
Normally, memoQWeb does not take snapshots of the documents, neither before nor after pre-translation.
To do this, use the settings in the Versioning section:
- Create minor version snapshot before pre-translation: Check this check box to create a new minor version before pre-translation. It is recommended to add a Note, too.
- Create minor version snapshot after pre-translation: Check this check box to create a new minor version after pre-translation. It is recommended to add a Note, too.
Run an analysis of the documents after pre-translation
Here you can choose how the analysis report that will run after pre-translation.
What happens with pre-translation if some segments are already edited, pre-translated, or confirmed?
If some target segments already have text, pre-translation overwrites them in some cases, and leaves them as they are in other cases:
- Empty (Not started) segments are always filled in if there is a match.
- Previously pre-translated segments are overwritten if memoQWeb has a better match than the one that was used earlier.
- memoQWeb will not change edited or confirmed segments. If you need to overwrite these segments, you need to clear them first
using the Clear translations command using the Clear translations command in the WebTrans translation editor.
When you finish
To run the pre-translation, and return to the Translations page, click the Pre-translate button.
To return to the Translations page without changing the documents, click the Cancel link.