building TM manually

babs guzman wrote:

OK, here's a case scenario: suppose you have a document of 1000 words with a lot of technical terms. You read the document but instead of starting to translate it, you'd rather first, look up all the terms you don't recognise and look up the equivalents in English. Second, when I've got all the technical term translations, can I type them manually into my existing TM, so that when I start translating my text, it will automatically recognise the terms that I have inputted and then put them also automatically into the target segment? !

Derek Gill Franßen wrote:

I don't work with Wordfast, but no I don't think it will work like that.

If it works anything like Trados (and I'm pretty sure that it does at this level), you have to differentiate between individual words (that go into a glossary) and segments (usually full sentences that go into the TM).

If you enter all of the terms into the glossary (not the TM), it will probably recognize those words and suggest the term you entered - not in the segment - but probably in some glossary interface (it's like your own dictionary on the side).

Individual words translated using Wordfast (which will put them into the TM) rarely do any good because it is usually sentences that you are translating, and it is sentences that are stored in the TM.

If the sentence you currently want to translate is similar to another sentence you have already translated with Wordfast (and which is then in the TM), Wordfast will pull up that old sentence so you can make any changes to match the new sentence.

Sometimes one-word entries do, however, help in a TM. For example "Address:" - once you've translated that, it is always the same after that and so your TM will recognize it. If the new sentence (segment in this case) is, however, "Addresses:" it might not be similar enough for Wordfast to suggest even "Address:" (you can change the values of how similar it must be).

This whole similarity thing may not sound important, but it is. A few different letters in two sentences of 10 words are enough to make them only 80% similar.

I'd suggest trying it out with sentences that are almost the same to experience what it does:

"The big, brick house on the corner is red."
"The big, brick house on the corner is blue."

Translate the first sentence using Wordfast. And then move on to the next sentence.

If the second sentence is similar enough (and we'll just assume that it is), Wordfast will suggest "The big, brick house on the corner is red." NOTE that it is just the first sentence (unaltered). That is what is in the TM from the first sentence. You can then just change the word red to blue, and that's it - you're done (that is what saves time).

After you've done both sentences your TM will have a translation for the first sentence and a translation for the second sentence in its database. You can experiment by translating a third sentence that is almost the same:

The big, brick house on the corner is green.

Depending on how similar this new sentence is to the other two already in the database, your CAT-tool will either suggest "The big, brick house on the corner is red." or "The big, brick house on the corner is blue."

The CAT-tool doesn't write or change the sentences but only regurgitates what you've fed it (with the exact same wording). It is merely a database of sentences you've already translated.

You don't even need the glossary, but it is helpful sometimes.

Even if you have entered the word "blue" into your glossary, it won't be inserted automatically into your new segment (they aren't that smart yet). But rather, you will probably see the translation for blue in your glossary and can either hit a button to insert it or simply copy & paste it into the segment (replacing the word red in the sentence that was suggested from the TM).

By using both features together you can actually come up with good results - especially after having worked with it for a while (the TM needs translated segments with which it can compare the new stuff and the glossary needs terms to recognize).

HTH

[Edited at 2006-08-07 00:30]

Milan Condak wrote:



You can Propagate terms from your Glossary 1, 2 and 3 and _automatically_ "translate" from source to target segment.

Term "translate" is reserved for segment = sentence. Translation units are stored in translation memory.

You can automatically "propapagete" words and expressions from your glossary, see:

http://www.condak.net/fast/cs/nula.html

You can "extract" all terminology before translating from your source text/s, translate with machine translation, edit manually and create reliable Glossary.

Next step is using the Companion for viewing terminology in your glossaries related to opened segment.

Milan
www.condak.net

There are two command in PB:
propagate and copy, the using of some combination of propagate + copying gives you automated "translating of terminology".
Copying + progagate = copying of "the translation" of recognised term in source segment into TARGET segment of the opended translation unit (TU). After Alt+Down is the whole TU putted into TM, not only certain term.

CAT recognise from TM a pair of segments = to source segment you get target segment.

The terms are added into glossary manually by double Ctrl+Alt+T.

Milan
www.condak.net

PS: "Building TM manually" is another cup of tea

babs guzman wrote:
I have a new document from a new client. I use an existing TM that's in the same field of endeavor or profession, say computers. I then choose TMComputer. I choose any glossary - that has a mixture of computer and medical terms. While I'm translating, will WF recognise terms that are in the TM and glossary and suggest them in yellow, or will it only recognise terms (in segments) from the TM? I guess this is the crux of the matter.
If you have time, I'd appreciate a reply. If you don't have time to spare, that's ok too. I'll just wade in the dark waters of Yves Champollion (LOL!).

Oliver Walter wrote:


I start my glossaries as tab-separated text files outside WF. To get WF to create a new one, there's the "Glossary 1" sub-tab in the "Terminology" tab. When you click that, you are invited to let it create a sample glossary or an empty one. As was written in another reply to your question, you can tell WF to add entries to an open glossary by using keys Ctrl+Alt+T.
While you are translating, WF is supposed to recognize whole sentences with a fuzzy similarity to the ones in your source text (using the TM). It's also supposed to recognize individual words (found in the source text) if it finds them in the glossary. How successfully it does this seems to vary for different users and to vary a bit with the different versions of WF (judging from posts to the Yahoo WF group).
Oliver