Although they have traditionally kept a low profile about it, IBM is not new to the translation software industry. IBM has been involved in computational linguistics and NLP (natural language processing) research for over 30 years and has a machine translation (MT) system called WebSphere Translation Server which has been around for about 10 years. You can try IBM’s machine translation online by clicking here.
IBM has not promoted this business aggressively and relatively little is known about this product. I searched the Internet for relevant customer case studies and found very few. Press releases about the product are also hard to find.
But IBM’s approach to language software will probably become more market-oriented and will result in new product offerings. IBM launched a project called n.Fluent in 2008 to develop their own machine translation system. One of the main reasons for this initiative was for security purposes, with IBM stating that in using online systems such as Google “you may unintentionally be sharing your Company’s Confidential information.” The MT was developed using IBM NLP technology and a corpora of over 36 million words translated by IBM employees in a worldwide crowdsourcing effort. As the crowdsourcing effort continues, this number is likely to be much higher.
In an interview with CNN last year Salim Roukos, IBM’s chief technology officer for translation technologies announced that the company’s intention is to take the machine translation project to market.
In my estimation, IBM will eventually release a new version of their translation server which will be based on a hybrid machine translation platform which combines rule-based MT [RbMT] and statistical MT (SMT). IBM can integrate their new translation server into their LanguageWare linguistic platform which offers a set of default language resources to support the processing of texts in more than 20 languages including Japanese, Chinese and Arabic. LanguageWare also offers a wide range of powerful NLP tools that customers can use to create a custom translation system for their own purposes. This includes tools for creation of dictionaries, rules and ontologies, dictionary look-up and fuzzy look-up, lexical analysis, language identification, spelling correction, normalization, part-of-speech disambiguation, syntactic parsing, semantic analysis and more.
IBM may be able to offer an alternative to current vendors of Enterprise MT software. By offering a building-block approach, customers will be able to build and deploy translation systems which are highly secure, powerful and customizable.
Related posts:
Comments