Start a new topic

A new possibility for CT !?

Hello everyone

I once talked with Igor about some of the features that my standalone machine translation package offers (unlike MS Translator and so on, it works off-line and can leverage user glossaries and TMs).

One of its unique features is what can be called "Auto-assembling of auto-fuzzy-corrected fragments."

This is done by designating corresponding elements of the source and the target as variables like this:

Source: One of its unique features is what can be called <$1>

Target: ...<$1>...

<$1> is a variable (my package uses <$1> to signify a variable) which is automatically replaced by whatever term or phrase is available from glossaries or memories (so long as a corresponding term/phrase pair is there).

This virtually corresponds to CT's fuzzy auto correction, I think.

Well, my MT package can do more; it accepts up to 10 variables in one translation unit like this:

Source: One of its <$1> is what can be called the <$2> of <$3>.

Target: ... <$1> ... <$3> ... <$2> ...

Note that variables can be in any order, so long as both segments have the same set of variables.

And, "auto-assembling of fuzzy-auto-corrected subsegments" combines two of these translation units (with variables auto-replaced).

Some MT suites call these translation units "sentence patterns or model sentences." Most standalone English/Japanese MT suites have this feature. Because English and Japanese are so different in terms of syntax, simply replacing variables (whatever they are called) in order of appearance rarely produces a good translation.

Well, as far as I know, this is the most advanced feature available from these packages (except for MT engines themselves).

Can this be a new possibility for CT?



Sounds very interesting. How about making this available as a plug-in? Or web search to a local resource.

Thanks Hans!

I really prefer a plug-in, if it's possible.

Because some of my agencies are very nervous about sending source segments to an external server, regardless of the security the operator promises.

Is this a technology you could/might want to share? I'm not against MT making my life simpler than it already is. I am against uploading my resources. So a local holy Moses would be greatianissimo.

Dear Hans

Well, I'm not a tech. All I want to do is to protect my customer's secrets.

To share or not to share? Maybe, if I can designate confidential parts of the sentence as variables, I would be willing to share it.

Hi Masato,

I like the idea! CT should provide a similar functionality soon.


Dear Igor


And one tiny request.

If this is possible, please consider providing a one-click feature.

My MT tool has one.

When you select a term/phrase as a variable in the current segment like this (underscored) with a dedicated button:

This is a sample to facilitate your understanding of CafeTran.

Then, two translation units are automatically created, as follows:

This is a sample to facilitate your understanding of CafeTran.

This is <$1> to facilitate your understanding of <$2>.

Well, I really thank you, Igor



I forgot to mention one.

When you select variables, before two TU's are created, the system displays the following information on the pane.

This is <$1=a sample> to facilitate your understanding of <$2=CafeTran>.

This helps me to confirm what I'm going to do.

Login to post a comment