| Translators rarely translate completely new documents that
are unrelated to any texts they have seen before. More often, a translator
may recall examples of translations for an idiom or domain-specific
expression but be unable to locate the source material to confirm her
suspicions.
In these situations, a memory of good translations
easily accessible by keyword searching would be an ideal aid to the
translator. CRL's ``Translation Memory'' tool provides this capability. In
most Translation Memory schemes, however, getting examples into the
database can be difficult. CRL's, ``XAlign'' provides the ability to
automatically pair sentences or passages from translated documents with
high accuracy. Translations can then be stored in Translation Memory
directly from XAlign, available for immediate searching for example
translations.
OperationXAlign and Translation Memory
are separate windows that work together. The first step is getting
translations into Translation Memory. Once translations are in, you can
then search for examples of past usages and quickly scan the examples for
the most appropriate ones.
You use XAlign to get translated texts from the
Tipster Document Manager. Texts are displayed side-by-side, applying a
segmentation strategy to chunk the texts according to a user-specified
scheme. This may be by punctuation, or it may be by SGML or HTML markup.
After segmentation, you can perform automatic alignment of the segments.
This is not foolproof, but it is often very helpful in getting an initial
pairing of translated segments. You can then manually change incorrect
pairings and send the results to Translation Memory to be stored in an
existing or new database.
Highlights
| After you have created a Translation Memory database, you
can then use the TM tool to search for examples. Some features of TM are:
In combination, XAlign and
Translation Memory provide you with the tools to manage translations and
make them available for future use.
ConfigurationXAlign segmentation
schemes can be designed by the user to meet specific segmentation needs. A
segmentation scheme for HTML documents that splits-up documents based on
HTML markup may not be appropriate for free-text, for example, and
sentence-splitting punctuation may not be the same between languages. In
Xalign, the segmentation schemes are transparently saved and available to
each user from session to session.
From within XAlign, you can also create new
Translation Memory databases and can then add translations to the
database. The databases are all managed by the CRL's NDS server possibly
on a remote computer. This frees up processing and indexing of the
translation texts from the local host computer. The location of the
Translation Memory databases is specified in the NDS configuration file.
StatusXAlign and Translation Memory are
integrated components of Oleada. They make use of the Tipster Document
Manager (TDM) and Norm Data Server (NDS) for fully distributed text
computing. To use XAlign and Translation Memory, you must import or create
translated documents within Oleada. You can then load them into XAlign for
alignment and save them to Translation Memory.
The automatic alignment algorithm used by XAlign was
developed to cope with real-world documents, including documents with
different markup schemes. At present, the algorithm is best suited for
French and Spanish, although XAlign is fully multilingual and alignments
can be prepared in any of the supported Oleada languages.
|
