Dr. Tom's Independent Software Reviews

SYSTRAN Premium 4.0

Introduction:
SYSTRAN (SYStem TRANslation) was founded 1968 in San Diego, California by Dr. Peter Toma, a a linguist researcher who worked on the U.S. government Machine Translation (MT) projects at Georgetown University developing Russian - English MT. The first SYSTRAN MT system was tested in early 1969 at Wright-Patterson Air Force Base in Dayton, Ohio, USA. Since 1970, the system has continued to provide translations for the US Air Force's Foreign Technology Division. In 1975, the Commission of the European Communities (CEC) contracted SYSTRAN to develop several European language pairs. Today 24 SYSTRAN MT systems are fully integrated into the CEC’s document workflow environment. In 1995, SYSTRAN PROfessional for Windows was released introducing the technology into the desktop tools market. 1996 saw SYSTRAN's use in SEIKO's hand-held translators and in 1997, BABELFISH was the first-ever online translation service, also powered by SYSTRAN’s technology after the AltaVista deployment . In its more than 30 years of history, SYSTRAN has an R&D investment measured in thousands of person-years. The resulting system is the Goliath of the MT world. SYSTRAN's dictionaries are enormous and diverse, it offers more language pairs than any other MT system (36) and has broad coverage of grammatical patterns across many text styles. SYSTRAN is developed under Linux and runs on IBM Unix, MacOS X and MS Windows. The Unix and Linux solutions have proven particularly important for SYSTRAN's enterprise deployments because Linux and Unix are not offered by other MT vendors, and these platforms are often used in large-scale server applications. SYSTRAN offers enterprise, professional, and personal MT products aimed at the corporate, small business, and consumer markets. SYSTRAN's MT technology integrates multilingual functionalities into information processing and exchange for applications such as eCommerce, CRM, Content Management, databases, corporate Intranets, email, Instant Messaging, SMS and WAP. SYSTRAN serves large corporations such as Ford Motor Company, Cisco Systems, NCR and Daimler Chrysler Corporation, Portals such as Google, AltaVista, AOL, CompuServe and Lycos, and Government and Public Administrations such as the European Commission and the US Intelligence Community.

Fig. 1: The SYSTRAN Toolbar - here shown within red outlines as the Add-In for MS Internet Explorer - is the primary user interface for all modules of the translation suite.
Double click on the image opens another browser window containing a Macromedia Flash Animation of the functionality of the Toolbar

Features:

SYSTRAN consists of the following components:

• SYSTRAN Toolbar: The SYSTRAN Toolbar is the primary user interface for SYSTRAN Premium 4.0. It is used to translate active files in Microsoft Word, Excel, PowerPoint, Outlook and Internet Explorer. One interface allows you to access machine translation with the same features in all of these applications. New in version 4 is the support of all recent Microsoft Office versions: 97, 2000 and XP, access to advanced linguistic options and display of alternate meanings (not available in SYSTRAN Personal). Besides the proprietary file formats of the office products (DOC, XLS and PPT), you can also translate RTF, MSG and HTML with full preservation of the documents' formatting. Maximum text size per file is 5 Mb. SYSTRAN 4's Microsoft Internet Explorer Plug-In adds support of version 5.0 and up. Web pages are also translated by preserving the formatting offering advanced linguistic options (New) and display of alternate meanings (New - both features are not available in SYSTRAN Personal).

Fig. 2: Creating a User Dictionary within the SYSTRAN Dictionary Manager:
please note, that the word form ("Category": e.g. noun, verb) of new entries is auto classified and they are verified against the custom dictionaries (checkbox and "Confidence" level) and can easily be assigned to every Domain defined for the dictionary.
Please click on the image for a closer look!

top

• SYSTRAN Dictionary Manager (SDM): SDM is a standalone application that helps you build your own customized dictionaries. The design strategy behind this great utility is to allow you to create dictionaries with the least amount of effort. It is not available in the Personal version!
  New in version 4 is SYSTRAN’s exclusive Intuitive Coding® technology which provides you with the ability to import pre-existing glossaries for usage during the translation process. Also new is the Bilingual or Multilingual (Premium Edition) User Dictionary Editor, a Headword marker and a confidence indicator for new terms. Now you can search your own local user dictionaries as well as search, browse and download Online User Dictionaries hosted by SYSTRAN. The Standard version allows you up to 2000 entries per dictionary.
  Version 4 of SYSTRAN Premium features additional new Expert Features such as the definition of general and custom domains and the support and editing of multilingual dictionaries containing up to 6,000 entries. In addition, you can now import and export Microsoft Excel and text file-formatted glossaries.
• SYSTRAN Translation Project Manager (STPM): STPM enhances the management of single translation jobs as well as overall projects. This allows you to keep track of all related essential information within a specific translation project, such as source and target files (Text Translation Duos, TTD), dictionaries, DNTs (do not translate: non translatable terms), as well as all other selected translation options.
  New features again include Advanced Linguistic options and the display of Alternate Meanings. Now you can include MS Word DOC, RTF, and plain TXT documents into your projects and translate with full connectivity to the Dictionary Manager module, side-by-side comparison and alignment of source and target texts, and highlighting of unknown words. Furthermore, translations corresponding to entries in your user dictionary can be highlighted and linked to the corresponding entry. New DNTs can be defined from within STPM without switching to the dictionary and a list of terms not found in the dictionaries can be generated including the frequency of such terms. This allows you a quick estimation of the work load involved with the project. Maximum text size is now 1 MB.

Fig. 3: Translation using the SYSTRAN Translation Project Manager and Text Translation Duos:
Please note, that SYSTRAN automatically adopts the formatting of the source to the appropriate term of the target document. Corresponding sentences are highlighted at the cursor position. Unknown terms are displayed in italic and red.
Please click on the image for a closer look!

top.

• SYSTRAN Multi-Translate Utility (SMTU): In version 3, SMTU was known as the Batch Translation feature. It is now a separate utility to ensure that batch translations are running in the background without any interference of your ongoing post-editing. You can batch translate any combination of the following file types: DOC, XLS, PPT, RTF, MSG, TXT and HTML and prioritize in which order the batch job processes your stack. A progress indicator displays the status of the single translation jobs and stores the times needed per document.
  New is the support of Advanced Linguistic options.

Fig. 4: Batch Processing using SYSTRAN MultiTranslate Utility
Please click on the image for a closer look!

top

• Finally the SYSTRAN Clipboard Taskbar (SCT) allows you to translate text up to 4 MB from almost all applications supported by MS Windows preserving the original formatting (SYSTRAN Personal only 5 KB). You can copy text from one application and past the translation into another with a minimum amount of keystrokes. This enables you to produce more translations in less time and increase your efficiency, which is especially important in real-time dialogs such as chat sessions and instant messaging.
  New features involve the "Active Clipboard" feature which automatically translates any text put into the clipboard without selecting the SCT first. A second novelty is again the support of Advanced Linguistic options in the translation process (SYSTRAN Personal excluded).

Fig. 5: The SYSTRAN Clipboard Task bar allows for the translation of any text within the Windows-Clipboard

top

SYSTRAN supports 36 bidirectional language pairs:
Western Europe: Between English or French and - German, Spanish, Italian, Portuguese, Dutch, and Greek (sold independently!).
Asia: Between English and - Simplified Chinese, Traditional Chinese, Japanese, and Korean.
Eastern Europe: English and - Russian.
Middle East: SYSTRAN announced the release of the English - Arabic language pair in 2004.

All SYSTRAN editions feature a pre-designed General Dictionary.
SYSTRAN Premium adds 20 Terminology-Specific Domains for context-sensitive translations, counting more than 2 million words:

• Business: Economics/Business, Legal, Political Science
• Colloquial: Colloquial
• Industries: Automotive, Aviation/Space, Military Science, Naval/Maritime, Metallurgy
• Life Sciences: Life Sciences, Earth Sciences, Medicine, Food Science
• Sciences: Computers/Data Processing, Electronics, Mathematics, Mechanical/Engineering, Photography/Optics, Physics/Atomic Energy, Chemistry.

SYSTRAN Standard and Premium versions 4 offer new Advanced Linguistic Options:

• Selection and ranking of the appropriate User Dictionaries and subordinated domains
• Language specific choice of the Imperative (infinitive, polite, ...)
• Language specific definition of the Personal Pronouns (formality, gender, ...)
• Display of terms not in the dictionary (Not Found Word marker)
• Marking and Linking of terms found in the User Dictionaries
• Exclusion of Capitalized Words from the Translation
• Preservation of the Textual Formatting
• Support of Document Sub Languages (Microsoft Word Add-In only!)

Editions & Prices:

New multilingual language packs (Italian pack, Portuguese pack, Dutch pack) to be published by the end of July 2003.

top

Usage & Evaluation

SYSTRAN 4 is a very convenient and complete MT solution. The integration of the SYSTRAN Toolbar into the most important applications of the MS Office Suite and the Add-In of the MS Internet Explorer offers fast and comfortable direct access from within the most frequently used production software. Users of alterative Office packages, Email-Clients and Internet Browsers can use the new SYSTRAN Clipboard Taskbar instead. If you tend to browse using a lot of simultaneously opened Internet Explorer windows, it is advisable, to deactivate the SYSTRAN Toolbar, as you otherwise risk a crash of your Browser or even of Windows. Another minor shortcoming is that I could not achieve the of display of alternative meanings in the translated documents.

SYSTRAN Dictionary Manager (SDM) is a great tool to build and manage your own user dictionaries. You can import already present bi- or multilingual dictionaries after pre formatting them to fit the required form. I initially had some problems to locate the information which specified the required format (Appendix C of the SYSTRAN Premium User Guide) as the Guides and Manuals are often not accurately cross linked. After working through the Appendix, import of even extended dictionaries worked like a charm.
The most advanced features of SDM are "linguistic clues". They are powerful means to improving translations. SYSTRAN's Advanced Coding supports the analysis of Morphology (e.g. Grammar, Gender and Number), Syntax (Prepositions), Semantics (Semantic Category such as "location" or "human - non-human") and Context in both - the source and target languages. An additional nice feature is the new option to browse and download Online User Dictionaries hosted by SYSTRAN and link them to your translations.
I initially did not understand how SYSTRAN computes the confidence level and how it detects an error (Fig. 2). The "Confidence Level" is a tool for efficiently reviewing a dictionary and focus efforts on parts that may need it. The display of an "Error" indicates a badly encoded entry. Both did not always seem to be reliable. After thorough reviewing and discussing this problem with SYSTRAN, it turned out that I used a quite unusual example which caused a rare grammar and semantic specific problem in the detection of both quality measures.
The German entry "auszugweise" can be either an adjective or an adverb. With his entry "auszugweise" = "in excerpt", SDM analyzes the German entry as an adjective but cannot guess the category for the English translation, thereby indicating an error. Again in this example, SYSTRAN is as good as MT gets today but it is still far away from producing linguistically perfect matches.

A very nice feature of SYSTRAN Premium which really give you a boost in productivity are pre-defined User Dictionaries supporting 20 Terminology-Specific Domains for context-sensitive translations, counting more than 2 million words. They also feature quarterly dictionary updates.

By its design, SYSTRAN Multi-Translate Utility (SMTU) is a fast and easy-to-use batch processing module now offering its own intuitively usable interface. However, I faced some serious problems freezing up the module and sometimes my whole PC: When translating, the registered applications (all MS Office applications) are loaded in the background and if they require any user feedback, e.g. because Macro security is not set to low, they can freeze or even crash. I verified this problem in MS Outlook, Word, PowerPoint and Excel. As it is not advisable to set Macro security to low by default, SYSTRAN should find other means to load and unload MS Office components. In MS Excel, an additional problem occurred: For each cell, translation would be larger or smaller than the source, a confirmation dialog is displayed.

SYSTRAN Clipboard Taskbar (SCT) adds support of almost all file types, MS Windows can display on your PC thus expanding SYSTRAN beyond the realm of only MS applications. Another plus is the preservation of basic formatting when pasting back the clipboard content into the same or even another application. However, the maximum Clipboard size of 5 KB in SYSTRAN Personal is too small to handle longer chapters of text, especially if a lot of formatting is included.

Overall Translation Quality performing a English to German sample translation of the content of one of the pages of SYSTRAN's own web site were quite disappointing (Fig. 3). This was the case when using generic settings comparable to those used in SYSTRAN Translation Services powering most of the Internet portals like Altavista's Babel Fish. Such translations use the default dictionaries and are on a "gisting" level. However, Translation quality did not improve tremendously, using SYSTRAN Premium and the appropriate specialized dictionaries only available in the Premium version. Given that I used SYSTRAN's own words to describe its functions which should be optimized for best results using MT, I am a quite disappointed with the quality.

In an initial installation problem involving the missing implementation of the SYSTRAN taskbar into MS Word XP, the SYSTRAN Support needed quite a long time to help me which resulted in own internet researches and several installations and de installations trying to solve the problem myself. As Software Reviewers usually receive preferred support, I wonder how long it would have taken them to successfully help the average customer. SYSTRAN Manuals and User Guides are well organized but need to be checked for missing or out-dated cross-references. Especially at a rate of almost 1000 US$ for the CD-ROM version of SYSTRAN Premium containing all language pairs available, I would have expected more printed documentation than just the 150 pages of the "User's Guide". Additional documentation is only available electronically in PDF or HTML format. The SYSTRAN Web site is a little poor and spartanic and does not offer a lot of background information or tutorials.

SYSTRAN Personal edition offers everything for basic MT needs at a very low price. Within the Standard version, especially the multilingual packs are quite moderately priced. The Premium versions are all overprized besides the CD-ROM pack which offers all of SYSTRAN for a good price.
Initial overall translation quality rates about 6/10 - after customizing and optimizing you might reach up to 8/10. However, some language pairs do much better than others. But even if all goes well, you should not expect results comparable to a well trained bilingual human translator familiar with the document's topic. This is, of course, not only SYSTRAN's failure but the general problem of all MT. In conclusion, despite a few shortcomings, SYSTRAN 4 is doubtlessly one of the best MT solutions available: MT by SYSTRAN is as good as it gets!

top

System Requirements:
Intel Pentium processor or compatible running Microsoft Windows 98SE / ME / NT4 with SP6 /2000 / XP; Microsoft Office Suite 97, 2000, XP; 64 MB RAM minimum and 40 MB disk space per language pair.

Company Information:
Headquarters: SYSTRAN S.A.; 1, rue du Cimetière, BP 7, 95230 Soisy-sous-Montmorency,
France; Tel: +33 1 39 34 97 97; Fax: +33 1 39 89 49 34.
North America: SYSTRAN Software, Inc.; 9333 Genesee Avenue, PL1, San Diego, CA 92121, USA; Tel: +1-858-457-1900, Fax: +1-858-457-0648.
URL: http://www.systransoft.com

Alternatives:

• HEISOFT Translation Engine (HTE): http://www.heisoft.de/volltext/heisoft/indexe.htm
• IBM WebSphere® Translation Server for Multiplatforms: http://www-3.ibm.com/software/pervasive/products/voice/translation_server.shtml
• PROMT/Reverso: http://www.e-promt.com and http://www.softissimo.com

top