[Corpora-List] Constitution

[Adam Kilgarriff]

Jean, Bart,

EU-multilinguality, with vast quantities of documentation and fast-growing
number of languages, is fascinating (and subcontracting to any human who
seems to have some relevant skills who happens to be available is a very bad
approach).

A particular issue is that there are many concepts in EU-speak which do not
exist outside EU-speak, so, prior to the existence of translations into
language X, there is no such thing as a correct translation for those
concepts.  Also, consistency and accuracy are important but naturalness
isn't.  (It is scarcely a concept that can be applied to EU-speak - who
would 'naturally' speak the prose of EU directives?)

A rational Brussels would devote its energies to developing a thorough and
complete lexicon, in all languages, of EU concepts, and then formalising
source language texts so that all instances of any of these concepts were
identified as such.  Then texts could be semi-translated, changing
concept-lexicalisations to the target language while leaving other words in
the source language.  All that would be left for either human or automatic
translation would be the grammatical glue to hold the concepts together.

A colleague on a course of ours from an accession state whose job was EU
documentation translation had, as a top priority, lobbying for further
copies of Trados Translation Memory for her in-house team, so that there was
some hope of them translating repeats of the same concept in the same way.
Let's not even think about her large team of freelancers...

I'm not sure that corpus methods are very relevant.  For new and newish EU
languages, the corpora don't exist yet or where they do, are likely to be
full of inconsistencies and not dependable.

Adam


-----Original Message-----
From: owner-corpora at lists.uib.no [mailto:owner-corpora at lists.uib.no] On
Behalf Of Jean Veronis
Sent: 15 May 2005 08:07
To: Bart Defrancq
Cc: CORPORA at UIB.NO
Subject: Re: [Corpora-List] Constitution

Bart Defrancq a écrit :

>
> - if I'm well informed, the Commission's SdT is not to blame for the 
> translation of the Constitution. Most of the work was done by the 
> Council's SdT. 

Well, thanks for the info, Bart. Then let's just add the Directorate 
General of Administration and Logistics (DGAL) of the Council to my 
reasoning.

> Unfortunately, there were no official translators for the languages of 
> the 10 candidate Member States (work on the Constitution was done 
> before the accession) 

Isn't this the core of the problem? Extension to new members states and 
8 new languges was known well beforehanhd. Couldn't Europe start an 
emergency plan to train and hire translators (instead it has *reduced* 
the translation effort!).

> and the translations were contracted out. The result pleads very much 
> against the "dismantling of translation bureaucracies", I'd say.-

That was irony ;-)

> the Académie française fiddled with the French (original) version even 
> after the text (one has to say texts) was officially approved by the 
> European Council. Some of the differences between versions may have 
> come from that.
>
Quite possible (if I find infos about this I'll certainly post it on my 
blog!). However, even supposing that each country had a need to massage 
the language of the text, there should be a final control within the 
European institution. After all, it's them and not the language 
Academies, who gave the green light to print it in the Official Journal 
of the Union in December 2004. This seems to confirm that something is 
wrong with the trnalsation and document production procedures and 
quality control within the European Institutions.

However, I am not sure that this debate belongs here. My message was 
more about using parallel corpora to spot possible translation errors. I 
don't know if this has been done before on other texts?

--j
  http://aixtal.blogspot.com