IBM Showcases n.Fluent Translation Service

According to David Lubensky, a computer scientist in IBM Research Division, the n.Fleunt project grew out of IBM’s need to conduct business globally. To that end, the company has been working on a machine-based approach to language translation that relies on a statistical model that translates words first into a set of numbers and then finds the corresponding set of numbers associated with letters and words in another language.

At present, IBM is using the technology internally to facilitate translation of Web sites into other languages and has created interfaces for its SameTime collaboration software and some mobile computing devices. The ultimate goal is to integrate the technology directly into mainstream applications. IBM also has ambitions to use this technology in interactive audio applications. To that end, the company has demonstrated the ability to translate newscasts recorded in one language to another, but the technology is not ready to be widely deployed for those types of voice applications.

In terms of language translation, IBM is in a race with a company called Language Weaver that also uses a statistical approach to translating languages. According to company CEO Mark Tapling, the quality of language translation software has dramatically improved as the algorithms surrounding language translation have improved and the cost of the processing power needed to achieve it has dropped.

Ideally, Tapling says the technology should be deployed in a scenario where it can be optimized around the lexicon of a specific vertical industry, but even the quality of translations will vary document to document.

Meanwhile, Google has already deployed a language translation service aimed at consumers and it’s widely expected to Microsoft will follow suit.

In general, language translation is getting closer to becoming a mainstream part of the enterprise application environment, which should serve to greatly facilitate global communications.