OASIS Mailing List ArchivesView the OASIS mailing list archive below
or browse/search using MarkMail.


Help: OASIS Mailing Lists Help | MarkMail Help

docbook message

[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]

Subject: RE: [docbook] Entities vs olinks

Even with English, you have to be careful with entities (or any
programmatic text replacement mechanism). Consider a case of a product
named "ACME server" that's rebranded as "FOOBAR server" and the sentence
"If you have installed an &ProductName; on the host..." The word "an" is
fine when &ProductName; resolves to "ACME server" but is wrong when it
resolves to "FOOBAR server". I know French and German have words that
contract in differnet ways depending on the gender of the noun, e.g. "de
la" v. "du" (instead of "de le"). I'm sure other langages have quirks
I've never imagined. For this reason and others we also preprocess docs
that we send to be translated. 

In addition to resolving entities and pruning <remark>s, comments, and
internal content, the xslt does some pretty printing. The goal is to
avoid a situation where the editor might introduce whitespace that might
confuse the translation tool and cause it to fail to match segements
from the translation memory. We also flag certain content (e.g.
programlisting, filename, database, code) as non-translatable. For
exceptions, the writer flags them as translatable before we run the l10n
prep script. This can dramatically reduce the word count and so
translation cost. The vendors seem happy to quote you a price based on a
word count of 10000 words and not make any allowance for the fact that a
sizable fraction of those words are in code listings that they won't
ever touch. 


> The most important reason why I think direct text expansion 
> *before* translation is the optimal solution is because the 
> translation of the entity text itself often depends on the 
> surrounding text. Phrases or words that can be used in 
> multiple contexts within English documents often won't make 
> sense in a translated document. If the entity has been 
> expanded to plain text *before* translation, then the 
> translator can alter the text as needed in each context.  But 
> if OLink or some other replacement is done *after* 
> translation, there can only be one replacement per entity, 
> regardless of context.
> Hopefully that makes sense. Sorry I didn't give more 
> background before...

[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]