Consider Subject and Language Being Translated Before Choosing Machine Translation Software, IDC Suggests
FRAMINGHAM, MA, AUGUST 21, 2000 – With the global omnipresence of the Internet erasing borders between countries, language differences remain one of the few inhibitors to doing business globally or communicating internationally. However, as automatic language translation software, commonly known as machine translation (MT) software, increases in popularity, language barriers may also become obsolete. According to IDC, machine translation is typically used to produce a real-time, rough draft translation that allows the user to understand text in a foreign language.
Exactly how rough is the draft translation and which systems deliver the best results? To find out, IDC put several machine translation products to the test. IDC tested European languages on the following subject areas: general information, technology, and finance.
"No one vendor has the definitive product in this area," said Steve McClure, vice president of IDC's Speech and Natural Language Software research. "Results vary considerably based on subject, language of the original content, and language the text is being translated into."
Collectively, the products IDC tested did the best converting Spanish to English. However, their worst performance was converting English to Spanish.
When choosing an MT system, IDC recommends developers look at more than the output quality. Other considerations should include performance, facility with specific text types — such as colloquial text — licensing structures, customizability, and economics.
"In any new application of MT technology, the developer will inevitably begin with questions about translation quality, asking what MT system produces the best output for a specific language pair or a special subject area," McClure said. "For several reasons, this question can be difficult for nonspecialists to answer."
· According to McClure the reasons for the difficulty include:
· The lack of a standard for measuring quality
· The challenge of quantifying human language (translators don't always agree on the best translation)
· The necessity of the person evaluating MT to have specialized knowledge of the languages involved, the linguistics, and the inner workings of translation systems
IDC's new report, Machine Translation Engines: An Evaluation of Output Quality (IDC #B22722), was written to help developers choose an MT system. The report evaluates the following systems: Systran, Barcelona, T1, Logos, and TranscendRT. To view the complete table of contents for this report, visit http://www.idc.com and search for 22722. To purchase the report, contact Cheryl Toffel at 1-800-343-4952 extension 4549.
IDC delivers dependable, high-impact insights and advice on the future of ebusiness, the Internet, and technology to help organizations make sound business decisions. IDC forecasts worldwide markets and trends and analyzes business strategies, technologies, and vendors, using a combination of rigorous primary research and in-depth competitive analysis. IDC provides global research with local content through more than 500 analysts in 43 countries worldwide. IDC's customers comprise the world's leading IT suppliers, IT organizations, ebusiness companies, and the financial community. Additional information can be found at http://www.idc.com.
IDC is a division of IDG, the world's leading IT media, research and exposition company.
# # #
All product and company names may be trademarks or registered trademarks of their respective holders.