This article is a report of a multilingual, team-based translation project of Windows Help files, carried out in an M.A. in Translation with Language Technology at the University of Wales Swanesa (U.K.). It comprises two sections: one is a procedure description for the project and the other is a discussion of Trados features revealed in the project. It covers both strengths and weaknesses using Trados and analyses the complexity in handling DBCS languages in Trados on the English-based Windows.
1. Setting up the project
irst of all, a folder called Trados Team 5 was created on Drive S, a shared drive on the University network, by the project manager (i.e., by the author), which consisted of various folders, each named after a Trados program (e.g., WinAlign, Workbench). Within them, subfolders for each language were created. The advantage of setting up the folders this way is to improve the general organisation of the team folder, and to avoid confusion between programs and/or languages. We decided to save all files on both Drive S and the Drive Q of each team-member to ensure data security, although version control problems could be a logical consequence of this kind of backup.
The source text (ST), called WinTest2, available on Outlook Public Folders on the University network, was downloaded by the project manager into the Team 5 folder so that the team members could access it at any time. Due to time limitations, the team was unable to meet and discuss the project workflow together, therefore a provisional workflow was sent by the project manager to all members by e-mail, providing an example of the procedure for this project.
To create a Workbench translation memory (TM) in Spanish:
Instead of moving an existing TM of WinTips1, WinTips2, and WinTest1 in Déjà-Vu, alignment of the ST and the target text (TT) was the chosen method for creating a TM because it is considered a typical workflow, as suggested in the Déjà-Vu User Guide.1 The TT of WinTest1 was edited in Word, following the corrections made by the translation tutor, using the Find/Replace command in the Edit menu. Then a new alignment project for WinTest1 was created in WinAlign. The same procedure was followed for WinTips1 and WinTips2. All the STs and the TTs were double-checked, and then the three aligned documents were saved as a project on Drive S and exported in a .tmx file, since that is a standard TM exchange file format. 2 From Workbench, a new TM was created with Multiple Translations unselected as default, which means that Workbench does not allow different translations for a duplicate source.3 That file was imported into the TM and the contents were checked in the Maintenance window.
It was necessary to translate all WinTips1, WinTips2 and WinTest1 into Japanese, since Déjà-Vu failed to recognise Japanese characters. To perform these translations, a preliminary bilingual MultiTerm termbase was needed to perform the translation effectively. Therefore, a termbase was created according to the International Word List on the Microsoft homepage. The entire list on the website was copied and pasted in a Word document, and converted into a table. The table was post-edited using NJ Star, since MultiTerm appears to be unable to recognise MS-IME (Microsoft Input Method Editors), unlike Workbench and WinAlign, which probably justifies several users' speculation that MultiTerm is not totally integrated into Trados's other programs and also proves that MultiTerm is still not Unicode compliant. The Mail Merge Wizard was used according to the procedure explained in the handout.4 Then the translations of three documents were carried out using this MultiTerm termbase and Workbench TM. Once a source segment was translated, the Set/Close Next Open/Get command (ALT [+]) was selected to send the target segment into the TM. There were some unknown words, which were manually added into the termbase. When finishing the translation, the same procedure as in Spanish (described above) was followed, using WinAlign to create a .tmx file, which would be used in ExtraTerm, as explained below.
To create a Workbench TM in Japanese:
To create a MultiTerm termbase:
The team created a MultiTerm termbase called Trados 5 and defined the termbase definition based on each language's characteristics, such as appropriate fonts for Double Bytes Character Set (DBCS) (e.g., SimSun for Chinese and Mincho for Japanese), part of speech, gender, etc. It was decided that only definitions in English should be inserted, because this is a localisation project and any term in a target language should signify the same as English. User IDs, passwords and input models were set up for team members, allowing them to edit only the necessary fields for their own target language, in order to avoid unexpected and erroneous editing of existing entries.
As a first step in populating the termbase, the alignment file saved in .tmx extension was opened in ExtraTerm and maximum two-word terms were extracted.5 Using the Concordance and also referring to WinTest2, important terms were validated, and then exported to a .txt file using the Export Wizard.6 The text file was easily imported into the team termbase by the terminologist. As for the lexicon of WinTest2, a monolingual term extraction project was created in ExtraTerm. Since ExtraTerm is able to recognise different word forms (e.g., print (v.) and printer (n.)), the most frequent word form in a term was selected depending on the context shown in the Concordance window. All new terms were validated and exported to the destination (i.e., the team termbase), then, finally, from the team termbase, they were imported by the terminologist. To search for missing entries in Spanish and Japanese, i.e., entries that were added by other team members, a filter was defined by adding Spanish and Japanese with criterion [!*] in the Filter Definition window. When a term with no Spanish or Japanese equivalent was found, an equivalent term was manually filled in, making full use of all keystrokes to facilitate adding and editing entries.
2. Performing the translation
Instead of translating WinTest2 from the beginning, the ST was analysed against the existing TM using Analyse from the Tools menu in Workbench. The advantage of doing this is that it creates the Unknown Segments in an .rtf file so only a small amount of the ST must be be translated. Each time the Set/Close Next Open /Get (ALT [+]) command was selected, the source and target segments were added to the existing TM. During the translation, unfamiliar terms were added into the team termbase manually (for Japanese) or using Copy to Clipboard (for Spanish).
Then WinTest2 was opened in a TRADOS 5.5 template in Word, and the Translate to Fuzzy was carried out. The Translate to Fuzzy command was selected rather than Translate in the Workbench Tools menu because it was obvious that the TM contained all the target segments and Workbench was able to perform all 100% matches automatically. Conversely, the Translate function is used to give freelancers access to the data that is contained in TM and MultiTerm according to Workbench Online Help.7
Then, a spell-check was carried out in Word. Although it took time for Word to search only for the target segments among many other source segments and segment delimiters and external tags, it was efficient to do it before the Clean Up because this command works both to produce the TT and to update the TM. Other small changes after translation were also made in Workbench using the Concordance in the Tools menu for spot editing, or using the Constraints in the Maintenance menu and the Source and Target Segments in the Global Changes window.8 Although switching between several windows could lead everything to go wrong and confining operations was not so simple, it was necessary to do it in order for the TM to be consistent.
3. Trados Features seen as strengths in this project
There are several strengths that are considered common to most Trados programs.
Various file formats are supported in Trados software, although in this project only a few different formats were used. In fact, different components exist for different file formats (e.g., TagEditor for HTML, PageMaker, etc.).
Network options are available in Workbench and MultiTerm to be used in large-scale projects involving many languages and translators, as exemplified in the user ID and password in MultiTerm. Workspace is a standalone application that is designed for project management and, from this, anybody can check the current procedure in the task list, and can also have access to other programs.
A program's settings can be freely defined by a user, e.g., the termbase definition in MultiTerm; and its display can even be changed, e.g., vertical or horizontal positioning in WinAlign. The structure and symbols in the setting of each component are for the most part commonly used (e.g., "!" means "not" in the settings of Export, Define Filter and Constraints in MultiTerm and Workbench); therefore, once a user has learned them in a program, s/he could apply them to other programs.
4. Features seen as weaknesses and difficulties encountered in this project
It is worth mentioning some specific points that arose during the project.
When translating the Unknown Segments, the text was presented out of the layout of the original document, which caused difficulty in choosing an appropriate translation. Especially in this project, the ST has a two-column part: "to do this" in the left column and "try this" in the right one, though in both sections the source-language sentences have the same structure: the imperative form. However, in Japanese it was necessary to differentiate the structure of both sections. During the Translate to Fuzzy, the translation had to be changed because it turned out to be wrong after the rendered translation was placed in the contextualised layout.
When performing a translation in Japanese, Workbench proposals were sometimes too complicated to choose from. Translating into Japanese seemed more complicated than into Spanish due to the difference of syntax between English and Japanese, which is wider than between English and Spanish. For example,
In this example, the Spanish segment reads in an order parallel to the English one, while the Japanese segment has a totally different word order. In Workbench, dictionary proposals appear just above the source-language word shown by a red line. Therefore, when producing a Japanese translation, the last dictionary proposal should be selected and put in the first place. Using keystrokes (ALT +[←][↓][→]) was a slightly complicated process and it was sometimes easier to render the translation directly, as if there were no dictionary proposals in the TM.
Trados has separate applications for different parts of a translation process: MultiTerm for terminology management, Workbench for translation memory, and Word for editing environment (the same format as the ST). This fact raised several difficulties as follows:
- Since Word is the editing environment, it is important to set up options appropriately in Word.9 Especially in the View tab of the Options from the Tools menu, field codes and other formatting marks should be properly selected so that hyperlinks and all hidden text in the Trados template can appear.
- It is also essential to use the same input methods, as well as font settings, throughout MultiTerm, Workbench and Word. In Spanish translation, some Spanish renderings were typed accidentally by English input method, since English input method was thought to be sufficient as long as the TT did not contain accented letters or ñ. Nonetheless, this caused confusions for Word in carrying out spell-checking, because some Spanish segments were considered as English and therefore Word offered some suggestions that were English terms.10
- Copying a Japanese entry from one program to another (e.g., from Workbench into MultiTerm) was not successful, though English and Spanish were done using the Copy to Clipboard. This is possibly related to the fact that both Word and Workbench work with MS-IME, whereas MultiTerm only works with NJ Star.
- To switch several windows and perform a command efficiently, all the keystrokes should be remembered and fully used, e.g., one keystroke leads straight to the command, while two or more clicks are necessary when using the mouse. Nonetheless, it should be borne in mind that careless keystrokes could lead to an unexpected command.
During the course of translation, these components (besides NJ Star for Japanese translation) were running at the same time, which resulted in some difficulties: too many screens to check,11 too much time to launch the programs, risk of malfunction in their interaction or even problems in the display, which could, in the worst case, lead everything to go wrong. Although the Fix Document command is available in Word, it was not always successful and it was sometimes necessary to reboot the computer. The starting order of the components is, therefore, fundamental and accidentally closing the window of a component must be avoided.
In this project, there were both time limitations and restricted resources. However, it highlighted some of the strengths and weaknesses of Trados in a practical team translation. For better team management, it would be useful to get access to Workspace, considering its characteristics in systematic management of all team members.12 This multilingual translation project, dealing with non-European languages, proved very complex in terms of the configuration of the languages of the programs.13 It was shown that Trados is a sophisticated but complex software, as discussed above. Although the potential of Trados was not fully realised in this project, a greater expertise of Word could contribute to overcoming some of the difficulties encountered in this project.
Atril UK Ltd., 2000, Déjà-Vu User Guide, East Sussex.
Microsoft, International Wordlist
(A comprehensive set of words and phrases that either appear in the Microsoft Windows user interface or are used in describing key concepts of the operating system to user data accessed in March 2003.)
TRADOS Ireland Ltd., 1994-2002, Trados Specialist Guide , Dublin.
TRADOS Ireland Ltd., 1994-2002, Translator's Workbench User Guide, Dublin.
TRADOS Ireland Ltd., 1994-2002, Translator's Workbench Online Help, Dublin.
1 See "Workflow" in Index tab in DV Help. The export (Spanish) should have been straightforward.
2 However, I could have edited the texts (WinTips1, WinTips2 and WinTest1) at once in Déjà-Vu editing environment, using the Find/Replace command, and exported the TM directly from Déjà-Vu then imported it to Workbench. It should be straightforward, but it should be borne in mind that the export procedure could result in a failure of data exchange.
3 Although depending on the context, the same target segment is not always appropriate as indicated in 2-5 in Trados Workbench User Guide, in this project the default settings was kept to maintain consistency in the TM.
4 SELM 02 Language Technology (2002-2003) Week 15: Strategies for Moving Terminology Data between Word, Déjà-Vu, and MultiTerm
5 I opted for this procedure because I did not succeed in moving data from Déjà-Vu terminology database in Spanish, and also to take full advantage of the .tmx file created by WinAlign.
6 See 9-18 in Trados Specialist Guide.
7 See "Translate command" in Index tab in Translator's Workbench Online Help.
8 See details in 3-5 in Translator's Workbench User Guide.
9 See the "Tips for Word" in 4-28 in Translator's Workbench User Guide.
10 However, it is still unclear whether Word determines the language from input locale/ keyboard layout.
11 e.g., one for verifying the TM (subwindows of ST and TT), one for editing in Word, and one more for NJ-Star.
12 I used Workspace in Demo mode offered by the Trados Demo CD; unfortunately this program was not available on the University Language Lab Network.
13 An example of this is in creating a Multilingual TM including DBCS. Although I tried all export formats (i.e., .txt and .tmx) in Spanish and Japanese, when importing them in a multilingual TM with appropriate settings, there were empty target segments found.