IBM Enters Machine Translation Space
Page 1 of 1
IBM Monday became one of the largest firms to tackle the task of translating languages for global Web communication with the launch of its WebSphere Translation Server.
Known as machine translation, an elementary use occurs when a user goes to a search engine such as AltaVista, types in say, "friend," and requests its proper translation in Spanish. If it's working properly, the user should see "amigo" as the correct answer.
But for Big Blue's purposes, the translation server meets needs on a much wider scale. The technology giant wants to sell the $10,000 server to enable enterprises to provide Web pages, e-mail messages and chat conversations in multiple languages, in real time.
Using the server, enterprises can create and distribute Web content globally, without the need for special Web pages or separate Web infrastructures. It also enables chats in multiple languages for both internal and external communications.
"The ability to support bi-directional translation for a wide range of languages was an important consideration for us, Stein said. "We also needed a system which actually understood the grammars of the various languages, much more than just a translation of individual words."
Steve McClure, vice president of Speech and Natural Language Software at research firm IDC, told InternetNews.com that machine translation is extremely cost-effective. And IBM's new product boasts a bonus in that it is bi-directional, which means a user can access text translation between English and French, German, Spanish and Italian, as well as from English to Chinese and Japanese and Korean.
"Machine translation allows for dynamic business content to be created at a fraction of the cost of professional translation services," McClure said.
McClure said until IBM's push into the sector, machine translation was attended to by about 60 various, yet mostly smaller players, including struggling Belgian firm Lernout & Hauspie and lesser-knowns such as Systran and Transparent Language.
McClure also noted that IBM's move will validate a sector whose software market is expected to reach $378 million by 2003. McClure said Moore's Law, which posits that the number of transistors-per-square-inch on a chip doubles every 18 months, also applies to speech recognition, the category in which machine translation is situated.
Running on NT, AIX and Solaris, the IBM WebSphere Translation Server can translate up to 500 words per second (speeds may vary based on processor speed and system configuration).
IBM's WebSphere Translation Server will be available in March 2001 from IBM, Lotus and Lotus partners.