Feature for translation memory.
It is about fine-tuning the leveraging of translation memories, project memories or terminology database.
Let's get started: When you create a project you typically select one or more resources to leverage. During word counting the system then uses these memories for pre-translation and for finding similar texts. How to you assign priorities to these resources? Which ones can be used for pre-translation, which ones not? Can I penalize this one memory? Can I give less weight to formatting differences? ... If these questions make sense to you then please read on!!
Prioritizing resources - "cascading memories"
This is not something new in Wordbee, but should not be missed out in this forum.
The order in which memories are referenced in the project is important. You tell Wordbee which memories to search first, second, etc. In the case of pre-translation, the first memory in the list, from top to bottom, that has a 100% hit "wins". So you likely would put your top quality memories in first position. The following screen shows an example:
To change order either click the up/down arrow buttons or use Drag & Drop.
This concept of prioritizing memories is also called "cascading memories".
Fine tuning - penalties and bonuses
Cascading memories are nice but sometimes a more granular tuning is required. Some scenarios:
- You want to disable a memory for pre-translation but still use it during the translation work.
- You want to penalize a lower quality memory and get only the most similar hits out of it
- You want to give a bonus to your top quality memory by letting through even the less similar hits
- You want to give less weight to formatting differences
All this is possible with the "Search settings" properties of each resource. In the screen above you can see this link to the right of each memory as well as the current project memory (see top right).
The settings look like this:
The upper half of the screen lets you enable/disable the use of this resource for pre-translation. Next to that it lets you give a bonus or a penalty for fuzzy matches and formatting differences (the inline tags). For example, a "Highest" penalty for text differences will let through only the most similar hits for this resource. The "Lowest" penalty (aka bonus) will show hits even with a very low similarity. The rule of thumb: Penalties are given to less interesting memories, bonuses are given to the top memories.
The lower half of the screen is about filtering certain types of hits where the translated text is identical to the source text and translations explicitly marked as erroneous. Erroneous segments are those where you had set the status of a segment to "red".
In the settings we make a difference between the scenario where you use the resource for leveraging or the scenario where you are editing the resource itself (for correction, translation, etc.). The latter case is most important when you translate a project. As you know, the project itself is a memory, more precisely a project memory.
While you edit a project, this project memory is used to find repetitions/translations within the all project documents. And here you might want to see erroneous segments in your memory hits...
Best practices and default settings
Best and recommended practice is to change advanced settings only if really required: The Wordbee defaults will be fine in most cases. Tuning is useful if you have very specific memories you want to promote with respect to others.
Another use case is where you want to globally penalise more or less text or formatting differences. For example, if you believe that inline tags should globally have less impact on the hit similarity. In such a case, please change the global default settings of your account:
Click "Settings" in the top navigation menu and then "Translation memories". You will get the page below. The global settings in this page will apply to all your resources by default. You can then of course override default individually per resource.