Termbase vs Translation Memory in Machine Translation

This Forum is for general issues about MadCap Lingo
Post Reply
minicyclist
Propeller Head
Posts: 60
Joined: Wed Apr 09, 2014 3:51 pm

Termbase vs Translation Memory in Machine Translation

Post by minicyclist »

Hello,

I haven't been able to fathom this from the Lingo user guide (which is woefully lacking in examples), but if I add terms to a term base, can they be used in a machine translation? I know the Translation Memory can be used pre-machine translation, and this begs the question, if I have a term base, can I add these terms to the Translation Memory? Generally confused, as to how to use these two different translation stores.

Grateful for assistance.

Ian
RobSwanson
Propeller Head
Posts: 21
Joined: Thu Apr 30, 2020 11:02 am

Re: Termbase vs Translation Memory in Machine Translation

Post by RobSwanson »

A Translation Memory is a list of phrases and its translations. If the entire phrase matches, it will be used to translate. Partial matches don't return.
A Termbase is a list of terms and words with the translation. It should populate the term from the termbase during the machine translation, be we've had issues with it. Footers don't translate, and sometimes it just doesn't translate it in the body of the text for some unknown reason.

From there, if you double-click the translated segment it goes to Translation Memory. Again, TM needs exact matches, TB finds individual words.
RobSwanson
Propeller Head
Posts: 21
Joined: Thu Apr 30, 2020 11:02 am

Re: Termbase vs Translation Memory in Machine Translation

Post by RobSwanson »

And... we have since learned more and my previous answer was wrong.

Translation Memory gets applied when activated. If your source segments match the Translation Memory, it will return in the translated segment. If you have it set for lower than 100%, it will drop any fuzzy matches in the window below the translations. Double click to apply it.

The Termbase is just a list of terms that does not get automatically applied. If the segment has one of the terms, in the TermBase window, you'll see it (though it doesn't always work, nor does it tell you where the wrong term is in the translated segment.) It also takes a long time to populate (several seconds, which doesn't sound long, but when you have to touch every segment, it is a long time.

What is missing is a Glossary. DeepL has a glossary function but it doesn't work with CAT tools like Lingo. A glossary is applied so that a given term in the segment is always translated with the word you have in the glossary. Where we are constantly struggling is with the word Plane, which often returns as airplane instead of geometric plane. A glossary would really help with this, BUT if you have a word that is used in more than one way, the glossary wouldn't work. For example, Key in English can be a keyboard key, or a key to a lock, or a main idea. These each translate to a different word in most languages.

We're hoping DeepL or another machine translation service gets a glossary to work with a CAT tool.
Post Reply