Translation process


win 1251
 



Translation process

From the very beginning, we were convinced that for high-quality translations and subsequent publishing (make-up) within deadlines it is necessary to create a process which will allow to computerize effectively all operations.

The process created consists of several sequential activities:

  • filtering a text and its preprocessing (preconditioning)
  • using a Translation Memory system
  • terminology development and dictionary customization
  • machine translation
  • text editing using a word processor
  • make-up of translated documents

Such sequential process allows to simplify essentially an operation of translators and DTP experts, to check and control a consistency of a terminology and style, and to computerize effectively all our activities.

Let's consider separately and more in detail each translation process step.

Filtering a text and its preprocessing (preconditioning)

We do not use DTP packages directly for translations. These packages are "bulky" and not effective for translations (for example, text processing, search and replacement features are limited). Therefore, before machine translation for each file following preliminary operations are performed:

  1. extracting a text to be translated using unique filters
  2. masking text parts not to be translated (for example, commands, messages, listings etc.)

Using Translation Memory systems

Since 1998, we started use successfully the TRANSIT (STAR), which belongs to Translation Memory systems. This package finds text segments in new files (as a rule, at a sentence level) which previously were translated already, and substitutes stored translations. Thus, translators edit new segments only. As a result, we actually "never translate the sentence twice" (the Trados' slogan).

Terminology development and dictionary customization

Terminology development and control are necessary to provide an efficient consistency management of terms and translation style. For this, following operations are fulfilled:

  1. preliminary analysis of used lexicon in source text and definition of terminology which is not available in our specialized dictionaries
  2. customization of existing or creation of new specialized dictionaries
  3. check of terminology used and dictionary customization quality during
    document postediting (after machine translation), and subsequent update of appropriate dictionaries

Machine translation of files

During this step, using machine translation systems from PROMT, file translation in batch mode is performed. Here it is important to select correctly the input/output format and to attach specific dictionaries in an appropriate order.

Text editing in the word processor

Correct selection of a word processor and its optimal application is the the most important factor for our translators. For their effective operation, such features are already supported:

  1. simultaneous multiple search and replacement (including regular expressions) in several files (no limitation for number of files)
  2. online operation with remote dictionary files (thus it is possible to search a term/word collocation by any mask simultaneously within several dictionaries; the search duration is maximum 2 to 5 seconds)
  3. synchronous scrolling of two files (source and target) in adjacent windows
  4. it is possible to put text fragments to several local clipboards

Make-up of translated documents

The publishing (make-up) operation is significantly computerized too. After the text is translated, it is inserted into a source file (using filters) at a proper place with saving all the formatting. Thus following operations are performed in automatic mode:

  1. checking whether such an insertion is correct
  2. font substitution
  3. operations specific to document format

Finally, following operations are performed (if necessary) in manual mode:

  1. correcting tables, figures, diagrams and all objects due to text boxes resized
  2. checking and editing of (automatically generated) tables of contents, markers, references, footnotes, abbreviation lists etc.
  3. checking a common structure of documents

After final make-up, the document is output according to customer demands (file archive, hardcopy, electronic publication)

Publications

Below you will find several documents which describe our translation process as well as documentation update features:


© "Argonaut" Ltd. 2002