Kevin Lossner wrote:
Wolfgang Jörissen wrote:
But it did not cross my mind in my last team project, either. We kept exchanging terminology databases just to able to track who added what. I would have preferred using the lexicon, because in that case, all we would have needed were the satellites.
In fact, it's the most logical solution if you want to pass a part of your project to a guy using DVX Editor...
Being the lone wolf that I am, I didn't think about team projects, but you're absolutely right. This could be a useful application.
The problem is the fuzzy matching doesn't work for lexicon.
If yours source language is heavily inflected, just like Polish, it's a pain in the arse, you must show some intelligence in order to make it work effectively.
Of course, you can send the glossary to a termbase but sometimes it's better to have a different (prioritary) lexicon.
However, my tests show that you can't add to a lexicon using the RTF or HTML table external view. You can only modify an existing lexicon. Of course you can get around this in a sense by making a lexicon with all the words in the project, but that includes a lot of useless crap.
You can reduce the Lexicon size i.e. strip the strings with more/less occurrencies/words than the level you find useful.
E.g., it makes no sense at all to use strings with only one occurence.
I almost always create the lexicon with all the words in the project and I start to work resolving them.
In this way, I get a pseudo-MT environment (with autoassembly).
For some closed and very predictible corpora, you may proceed up to 8000-10000 weighted words daily.
Of course, the third or fourth day, you're dead
[Edited at 2009-01-18 09:56 GMT]