Translation memories with multilingual target languages are supported and all TMs can be used bidirectionally.
Create New Translation Memory
In the Memsource Cloud app, you can create a new translation memory by selecting Translation Memories in the left-hand navigation and then clicking New. Alternatively, if you want to create a Translation Memory within a project, access the project Dashboard and select Create New in the Translation Memories section.
- Enter TM Name
- Select a Source Language
- Select one or more Target Languages
- You can select (if applicable):
- Business Unit
- Hit the Create button
Translation Memory Search
Use the Search box to search for a word or phrase in the TM (wildcard search is supported).
The results can be edited by double-clicking in the text area or deleted with the xbutton.
The columns with individual languages results can be re-arranged using drag&drop by mouse.
Translation Memory Import
The TMX file has to match the source language of the translation memory and at least one of the its target languages. When a TMX file includes multiple target languages, only those present in the translation memory will be imported. The TMX file version should be 1.4, although Memsource should be able to import other versions as well. The size limit for a TMX file is 1GB.
TM Import Options
- Remove all non-Memsource tags
- Tags that cannot be converted to Memsource tags will be removed (if you are not sure about this option, just leave it selected)
- Strict locale matching
- When this option is selected, Memsource will apply strict matching to language codes between the translation memory and the TMX. For example, if your TMX file includes both "en_gb" and "en_us" and your TMX file includes "en" and "en_gb", you may want to select this option. It will ensure that "en_gb" in the TMX will be imported into "en_gb" in the TM and "en" in TMX will not be imported into the TM. When this option is not selected, the locale matching is not strict and consequently if you have "en_gb" in your TMX and "en" in your TM, "en_gb" will be imported into "en" in the TM during the TMX import.
After the import, segments will be deduplicated. This is the Memsource definition of a duplicate segment:
- The source and target of the two segments are identical
- And also the context of the two segments is identical (the context = the neighboring segments)
When segments are identical (including their context) but their translation differs, the translation in the TMX will overwrite the existing translation in the translation memory.
TM Import Error
At times you may receive an error during a TM import. The single most frequent reasons for import errors are not corresponding languages, illegal characters or other irregularities in the file which you attempted to import. If you know how, correct the file and then reimport, otherwise email the translation memory file to Memsource Support, we will fix it for you and send it back, so that you can re-import it.
SDL TMX Import
TMX exported from SDL Trados is supported in Memsource, but only certain attributes will be imported into Memsouce TM:
- 'creationdate' will be converted to 'created_at'
- 'creationid' will be converted to 'created_by'
- 'changedate' will be converted to 'modified_at'
- 'changeid' will be converted to 'modified_by'
Important: Attributes 'created_by' and 'modified_by' are connected to the Memsource Users database and only Memsource Usernames are supported. See more details in Translation Memory Metadata.
MXLIFF files can also be imported into the translation memory via the Import TMX button. This import feature has been primarily designed for emergency situations when, for example, it is not possible - for whatever reason - to export a translated file from Memsource. However, reimporting the original source file and exporting it works. In that case - to utilize the already available translation - a user can pre-translate the reimported file from the translation memory. This works well, except for segments which were split or joined during translation.
The MXLIFF into TM import therefore attempts to revert the splitting and joining in the MXLIFF file prior to its import and in this way makes restoring a translation much easier.
Excel files (XLS and XLSX) can be directly imported into the translation memory. There are two ways in which you can import TM data: Without IDs or with IDs. The size limit for Excel file is 1GB.
XLS Import without IDs
The format of the Excel file should be a plain list of segments with the language code at the top. Make sure the languages listed in the XLS correspond to the languages in the Memsource translation memory into which you plan to import the Excel file. Use only supported language codes to identify the languages in your Excel file. When importing without IDs, new segments will be created. However, duplicates within xlsx will be removed. Please note that segments translated in Memsource are saved to TM with Context (Prev and Next). Segments will not be de-duplicated unless they have the same context.
XLS Import with IDs
In this case the Excel file includes not only segments but also IDs. This import method allows for two additional options:
To delete a segment, add |delete after an ID.
To update a segment, add |update after an ID.
Independently of whether the |update option is used, newly imported segments are still deduplicated based on their context. For this reason, it is advisable to always export and import segments including the previous segment and next segment.
Translation Memory Export
Translation memories can be exported by clicking on the Export button and selecting the file format method: TMX or XLSX.
One can select which target languages are to be exported for multilingual translation memories.
To export a TM into a TMX file, hit the Export button and select TMX. In Multilingual TM it is possible to select target languages for export.
To export a TM into a XLSX file, hit the Export button and select XLSX. The XLSX will include several columns: ID, source language, target language(s), previous/next segment, etc. You can specify what attributes you wish to export. If you plan to edit the TM and then import it back to Memsource, we recommend exporting all the attributes.
To launch the alignment dialogue window first select a TM with the same source and target languages, then click on the Align button and select between the two options:
- 1 source + 1 target file
- Multiple files
The Multiple files option is useful for aligning multiple files at once. The file pairs (source/target) must have identical names and are imported in a zip file with folders named "source" and "target" containing corresponding files.
The output of the alignment is a XLSX file with three columns: segments in the source and target language and the filename of the target file. The file can be further edited in MS Excel and later directly imported into the Memsource translation memory.
The aligner supports these file formats. Custom segmentation rules can be selected for the source and/or target files for alignment.
Creating a separate translation memory is recommended for the aligned segments (it could even be called Aligned TM). In this way one can easily distinguish between the segments originating from the alignment and the actual translation.
Translation Memory Metadata
The following metadata is supported:
- Created date/time
- Created by (only Memsource Usernames)
- Last modified date/time
- Last modified by (only Memsource Usernames)
- Context (Prev and Next, or Context (Segment) Key)
To ensure that metadata is correctly imported from a TMX, the TMX has to have the following metadata format:
Properties in source TUV element:
<prop type="context_prev">Text of the previous segment</prop> <prop type="context_next">Text of the following segment</prop>
or - for context based on a segment key:
<prop type="x-context_seg_key">Context Key</prop>
Properties in target TUV element:
<prop type="created_at">1322746823589</prop> <prop type="created_by">Some name</prop> <prop type="modified_at">1323854662890</prop> <prop type="modified_by">Some name</prop> <prop type="project">Project name</prop> <prop type="client">Client name</prop> <prop type="domain">Domain name</prop> <prop type="filename">File name</prop>