Start a new topic

Total Recall and Project properties

This has escaped my attention:

"The Project Configuration panel allows you to set some properties for the current project such as Subject, Client, Context etc. Those properties propagate automatically to the segments added to the translation memory during translation. When you store the translation memory in Total Recall, the properties are also maintained in the corresponding property columns of the Total Recall table. This makes it possible to recall only the segments with a defined property or set of properties from Total Recall in another translation project."

This means that setting project properties is mandatory for Total Recall to work properly? And what happens if project properties are not set?

Hello Mario,

No it isn't mandatory at all. If no properties are set, no properties can be filtered, but's all. Filtering properties is an advanced feature, not a requirement.

But it could be good practice to define such (metadata) properties in your project memories, if you think they will be relevant to you in the future.

For example, you might want to have Total Recall not only recall those segments that have words found in your project, but which belong to a specific Subject, Client, Context, etc.

This is not only true for Total Recall, but in general for Translation Memories.

If you click the Filter button in the TM options of any translation memory (including one that is created from a Total Recall table), you can further filter the TM with such properties.

The options are explained here: The file also explains the additional options found in the Dashboard's Total Recall TM options)

The difference, is that in addition to Total Recall > Recall memory, you can also open the Total Recall table itself.

TM properties can be added or removed even after the fact, when editing a TM or working on a project. This is done in Task > TMX memory > Set property for TMX units and Remove property from TMX units.

To edit a TM, [select Dashboard menu button > Project type > Edit translation memory], and drag and drop the TMX onto the Dashboard, then select Edit. All standard resources, Filtering, Find and Replace and Tasks are available to you.

If, in addition to any other TM, you still make use of the standard Project TMs CafeTran offers to create, what I think is important (but still not mandatory), is to make sure all TMs are up to date when you finish a project, and that these are renamed, stored and organized in a meaningful way to you (including setting TM properties).

TMs inherit the project properties, so it is better to set these at project creation, but this can also be done at a later stage, as explained above.

This way, you can be sure to be able to easily create ad hoc "Big Mama" TM joining many TMs in one (Memory > Open memories folder), update or recreate a Total Recall table at any time.

Each month, I create a renamed copy of each ProjectTM.tmx (example: 20180708

Client_Six letter code_Project name_Language pair) and store it in the corresponding language pair TM folder.

I wouldn't rely on just storing my TMs in Total Recall. Make sure you keep them organized in a way that will let you quickly reuse them in the future, no matter the program that you will use, and including a Big Mama approach or a Total Recall approach.

I have never used Deja Vu for translation, but I think this concept comes from there, see woorden's blog post on that:

For example, woorden does not define properties at all, I think.

Will project and TM properties (metadata) be relevant to you? This is the question that you should be asking. If the answer is Yes, then you should take care of this too.

Hope it helps!


1 person likes this

How to make sure a Project TM is up to date in the end of your project?

Two methods:

- Only if you already have a ProjectTM.tmx attached to the project, and if that ProjectTM uses the option "Keep newer duplicates" = Go to Memory > Import > Import segments from projects. Then save the memory.

- This works even if you don't have a ProjectTM: Make sure no segments are filtered (unless you want to store a subset of the project segments to a TMX), then Project > Export and exchange > To TMX. Give it a relevant file name, and you're done.

This is also a way to easily create a Project TM from older projects, if they don't have one (or if they are external projects).

1 person likes this

Hello Jean,

Neither I have ever used project properties in Dejavu, essentially because I never followed the classic Big Mama path. But, it may change with Total Recall, although I don't know whether it would make fully sense because—as far as I think I have understood so far—Total Recall also works at sub-segment (fragment?) level, making it sort of property-agnostic since fragments can equally apply to completely different projects. 

In other words, Total Recall might be more universal, and perhaps more useful, without setting any specific property, in particular for Japanese source texts due to the lack of inter-word spaces. 

And last but not least, I don't use auto-assembly because I find it more of a hindrance than a help, especially when translating from Japanese.

Certainly I never delete project memories even after sending them to Total Recall.

Hello Mario,

If you don't need filtering, it's really fine not to use it in CafeTran, just as I understand you did in Deja Vu.


Although I haven't used properties filtering much so far in Translation memories or Total Recall, I tend to store information such as Client, Subject and Project name.

Filtering is interesting if you wish to limit the TM results to those that are relevant for the specific project you are working on.

Example: If you have translated material for Toyota and for Honda, maybe you want to make sure that you use the manufacturer's or client specific terminology. Same goes if you translate some subjects where getting TM matches from other fields might lead to incorrect terminology or phraseology.

It is easy to set these properties at project creation (except for external projects), but since you store your project memories, you can have Total Recall to create a general (low priority) TM and you can still create a specific, ad hoc TM from any collection of TMs that you wish (if you store the name of the client, for example, you can easily create a Client-only memory [Memory > Open memories folder])


So, the interest of filtering does not have lie in sub-segment matching, as I understand it.

With Total Recall, you get : 

  1. a (database) table, that you can open and search it instantly as is (Total Recall > Memory Tables > [choose the table you want]) 
  2. a way of creating a normal Translation memory (TMX) for your project, where you recall all or (most probably) a subset of segments stored in TR.

If you click the gear icon in the Total Recall section of the Dashboard, you can set the Total Recall preferences for the TMX that is created for you. As you see, almost all settings are the same as for a standard TM:

Only one setting is added:

Recall in context (hits per word) = Checkbox and number field. This value sets the linguistic filter which analyses the current project, bringing back only the segments that contain words of the source document present in the project. The value determines the maximum number of the segments which are to be recalled for one word of the source document being translated. Checking out the “Recall in context” box lifts the filter completely and recalls all segments from the Total Recall table to the working memory.

Default value: 1000.

So, you can still recall ALL segments, if you remove uncheck this option. Depending on your Total Recall, you might want to set this value higher, but this will create a bigger TM. In that case, definitely use Preliminary memory matching option in the Workflow integration.

This is also explained here:

To understand the difference between Workflow integration >Automatic and Preliminary memory matching, you can read this:



I find the concept of Auto-Assembling very nice and worth trying. It might require fine-tuning your Preferences, assigning priorities and setting some resource to "Keep out of auto-assembling" to work correctly. I too haven't been able to take advantage of it in most of my projects. I think some languages are better suited for this method, but it also depends on the types of projects or resource that you use. For example, I had better luck with content that had many repetitions and only slight variations.



I really don't know much about languages where words are not delimited by spaces, but here's one setting that you might want to try, or make sure it is applied to your TMs if your source language is Japanese:

Matching type > Fuzzy without word separator = With this option, CafeTran analyzes source segments on a character basis, which is suitable for languages without a word boundary (e.g. Chinese or Japanese).

Also explained here:

Login to post a comment