Información sobre el sistema de ayuda a la traducción DÉJÀ VU
Carlos Subirats Rüggeberg
Carlos.Subirats at UAB.ES
Wed Jun 23 11:43:30 UTC 1999
INFOLING Lista moderada de lingüística española
http://listserv.rediris.es/archives/infoling.html
http://www.rediris.es/list/info/infoling.html
Estudios de Lingüística Española: http://elies.rediris.es/
Envío de información: infoling-request at listserv.rediris.es
Editor: Carlos Subirats Rüggeberg <Carlos.Subirats at uab.es>
Colaboradoras:
Paola Bentivoglio <pbentivo at reacciun.ve>, U. Central Venezuela
Mar Cruz <mcruz at lingua.fil.ub.es>, U. Barcelona
Eulalia de Bobes <ebobes at seneca.uab.es>, U. Autónoma Barcelona
Emma Martinell <martinell at lingua.fil.ub.es>, U. Barcelona
Rosa Ribas <Ribas at em.uni-frankfurt.de>, U. Frankfurt
____________________________________________________________
Información sobre el sistema de ayuda a la traducción DÉJÀ VU
De: http://www.atril.com
http://www.atril.com./whatsdv.html
Con información proporcionada por X. García,
Ampersand Traducción Automática S.L <xavi at ampersandsl.com>
____________________________________________________________
Indice del contenido de este mensaje:
1. Características generales de Déjá Vu
2. What's Déjà Vu
3. Déjà Vu prices: product description and prices (in euros)
4. Overview of CAT (Computer Aided Translation) concepts
----------------
1. Características generales de Déjá Vu
Déjà Vu es un programa de traducción asistida por ordenador,
que pertenece al grupo de los 'programas de memoria de
traducción', puesto que efectúa una pretraducción de un texto a
partir de una base de datos, que está formada por el corpus de
traducciones realizadas previamente por el traductor. A
diferencia de otros sistemas, Déjà Vu reutiliza frases
traducidas anteriormente o extrae equivalencias multilingües a
partir de los textos activos.
Además de esta capacidad de acumulación y extracción de
información, Déjà Vu posee otras prestaciones, como p. ej.:
- la utilización de filtros para varios programas de
autoedición,
- un alineador de corpus,
- un gestor de terminología, etc.
Déjà Vu puede tratar el alfabeto griego, el cirílico, e
incluso lenguas como el árabe o el hebreo.
Los límites de Déjà Vu son quizás la ausencia de
"language-dependant components", es decir, componentes que hagan
más "inteligente" la extracción y reutilización de elementos de
la memoria. Por ello, Ampersand SL tiene como proyecto
incorporar componentes lingüísticos para diferentes lenguas,
como p. ej., analizadores de constituyentes, etc.
La empresa Ampersand, que produce este programa, organiza
presentaciones gratuitas del programa en sus oficinas. Para
poder asistir a dichas presentaciones, debéis poneros en
contacto con Ampersand SL:
Xavier García <xavi at ampersandsl.com>
2. What's Déjà Vu
Déjà Vu version 2 is an industrial strength, yet user
friendly, powerful and customizable CAT (Computer Aided
Translation) system.
Basically, Déjà Vu works by storing your translations
in a "memory database" and instantly retrieving that information
whenever it finds something equal or similar to the sentence you
are about to translate.
Déjà Vu includes TermWatch, a fully integrated terminology
management system, as well as a File Alignment Wizard that helps
you create memory databases from translations you have done in
the past.
Déjà Vu Interactive (DVI) is the program you use to actually
translate your projects, and to take advantage of Déjà Vu's
large range of time-saving features, some of which are:
- You can ask DVI to pretranslate a project; i.e., to
search the memory database for all the sentences in
your project. Any matches found, both exact and fuzzy,
will be inserted in the correct places. Imagine
translating a second edition of a past job: you will do
it in a matter of minutes.
- Assemble, our newest technology, takes a closer look
at the memory database than Pretranslate, and in many
cases succeeds in assembling a translation for
sentences where Pretranslate has failed.
- Propagation allows you to propagate a new translation
(which was not in the memory database when you
pre-translated the project) across all the files in a
project. Fuzzy propagation (to similar sentences) is
also possible.
- Scan the database at any time for a source sentence,
or a portion thereof. Typical scan times range from a
few milliseconds to a few tenths of a second.
- Look up a word in the TermWatch terminology database
with just a mouse click.
- Build the project lexicon and resolve it using the
memory database or the terminology database.
- The Learn function intelligently evaluates source and
target sentences in the memory database to give you the
most probable translation for an unresolved term in the
lexicon, or for an untranslated expression in the
source text.
- Spell checking for US English, UK English, German,
Spanish, Finnish, French, Italian, Dutch, Danish,
Brazilian Portuguese, Norwegian (Bokmal) and Swedish is
included. New dictionaries can be created.
- AutoText. Similar to Microsoft Word's AutoText.
Global and project-level entries can be defined.
- Project sentences can be ordered alphabetically (so
that you can spot and translate similar sentences at
once) or by their normal order. You can place bookmarks
in the document to be able to instantly return to
specific places.
- Sentence locking allows project managers to
explicitly prevent the translation of specific
sentences.
- Count words and characters per file or per project,
both source and target, with project and file totals
(these can be saved to be used for invoicing).
- Extensive project management options which simplify
the completion of large multifile and multilingual
translation projects. Not only can several translators
work on the same project in different languages, but
several translators can work in the same language on
the same project. Project managers can distribute
'satellite' projects and receive/distribute new
database entries as separate text files.
- DVI can build and handle projects for Word, RTF,
FrameMaker, Interleaf, QuarkXPress, HTML, HTML Help,
Trados WorkBench, IBM TM, RC, C/C++ and plain text
files.
There are utilities for maintaining memory and terminology
databases. All databases can be reversed, so that if you have,
for example, a database for English>German, you can create one
for German>English.
Déjà Vu's ability to handle very large databases makes it
equally suited for individual translators and translation
companies.
If you run into any problem, you can take advantage of Déjà
Vu's unlimited free technical support.
New features suggested by users are normally implemented in
a few hours, or at most in a few days.
Déjà Vu runs under Windows 95 or Windows NT Workstation. For
optimum performance, a Pentium processor with 32 Mb RAM is
recommended.
3. Déjà Vu prices: product description and prices (in euros)
DV2FF, Déjà Vu, dongle-protected, with alignment
utility, TermWatch and support for FrameMaker,
PageMaker, QuarkXpress, Interleaf, HTML, HTML Help,
SGML, RC, C/C++, RTF, Word, PowerPoint, Plain text, IBM
TM, Trados WB.
First license 990
Licenses 2nd through 5th (each) 792
Licenses 6th through 10th (each) 693
Licenses 11th- (each) 594
Special discounts available for universities
Please write to sales at atril.com for details.
DV2EV Déjà Vu Evaluation Kit. Includes
CD-ROM, printed manual, and 30-days
license 50
4. Overview of CAT (Computer Aided Translation) concepts
Computer translation systems can be roughly divided into two
distinct categories:
- Machine (or Automatic) Translation (MT) and
- Machine Assisted Human Translation (MAHT) systems.
There are as many synonyms and subtle definitions as there
are CAT researchers and tool vendors, but there is a clear
distinction between the two categories: the use of linguistic
in-formation.
MT systems produce translations by applying linguistic
knowledge; they linguistically analyze the source language text,
map linguistic structures to their target language counterparts,
and construct a translation from scratch.
Translation Memory (TM) systems, the main type of MAHT tool,
store previous translations and offer these translations when
they encounter identical or similar sentences when translating
new material. TM tools learn while you use them, so the more you
use them, the more useful they become.
MT systems have two distinct disadvantages. They are
entirely language dependent, since they must employ advanced
linguistic techniques that apply exclusively to a specific
language. This means that MT systems are limited in the
languages they can handle. In addition to this, since a computer
program lacks an understanding of language, context and
innuendo, MT systems produce sterile and often awkward
translations. Most MT systems have dictionaries and phrase
patterns that can be customized, but they can never match the
quality of professional human translation.
TM tools, on the other hand, do not translate by themselves
at all. A translation memory is a database in which source
language sentences are stored together with their translations.
When the TM tool is about to translate a new sentence, it
searches the translation memory database for the sentence it
wants, and presents the user with the translation(s) of the same
sentence or similar sentences which the translator did in the
past. While in many cases they do not provide a translation per
se, providing the translator with similar material helps to
ensure the consistency of style and terminology by allowing the
translator to easily consult previous translations.
In other words, MT systems translate but lack style, while
TM systems do not translate but provide consistency.
Where does Déjà Vu come into all this? Déjà Vu (DV) goes one
step beyond traditional TM tools. While other tools only display
exact and fuzzy matches (matches for similar source language
sentences), DV intelligently uses the memory database,
terminology database and project lexicon to increase the quality
of fuzzy matches by working with smaller segments and words. For
close fuzzy matches, DV will often produce a perfect translation
in your own style. If you specialize in technical translations,
you may well find that DV doubles or even triples your
productivity.
However, DV is not a mere translation TM tool. It is a
full-featured translation environment with many additional
features that will increase your productivity and the quality of
your work. Whatever type of file you work with, DV presents you
with a uniform interface, displaying only translatable text, so
you can work comfortably without having to worry about
overwriting formatting and layout information. DV includes
filters that allow you to translate Word, RTF, PowerPoint, SGML,
FrameMaker, Interleaf, QuarkXPress, HTML, HTML Help, RC and
C/C++ files. Support for Trados Workbench and IBM TM2 files
ensures that even when your customers work with other tools, you
can still use DV.
DV provides extensive project management features which
simplify the completion of large multifile and multilingual
translation projects. Not only can several translators work on
the same project in different languages, but several translators
can work in the same language on the same project; advanced
features such as sentence locking, project lexicons and
configurable project security make this a breeze. Project
managers can distribute satellite projects and
receive/distribute new database entries as separate text files.
Powerful statistics options allow you to quickly produce quotes
for your customers, to evaluate the progress of a translation,
and to accurately bill your customers once the translation is
completed.
DV also includes a full-featured terminology management
module called TermWatch. Term-Watch terminology databases are
integrated into the DV translation environment, but Term-Watch
can be used as a standalone tool. Configurable, system-wide
hotkeys allow you to access the terminology databases from
within almost any Windows application.
To get you started quickly, DV includes an alignment tool
which helps you create memory databases from translations you
have done in the past.
----------------------------------------------------
Normas para el correcto uso del correo electrónico:
http://www.rediris.es/mail/estilo.html
----------------------------------------------------
More information about the Infoling
mailing list