ISEdb.COM

Earlier today, Google announced in an official blog post the release of Google Translator Toolkit, a new Web application that builds on the popular Google Translate that might facilitate the growth of sites like Wikipedia in its alternate language versions.

“While we think Google Translate, our automatic translation system, is pretty neat, sometimes machine translation could use a human touch. Yesterday, we launched Google Translator Toolkit, a powerful but easy-to-use editor that enables translators to bring that human touch to machine translation,” Google Translator Toolkit members Michael Galvez and Sanjay Bhansali posted on the blog.

As the two later explained, the application has the ability to learn adaptively to find the best possible translation for a string of text into another language, and also includes a functionality that will allow for rapid translation of Wikipedia pages. As users test the platform they can correct mistakes, add context and improve the translator’s database, from which the tool can adaptively learn new phrases and sentences.

Moreover, previously translated sentences can become translation suggestions for future users after being rated by a sufficient number of users. One good reason to share translations with others, Google explained, is so that they can be reviewed for consistency and style. With time and a sufficient number of user inputs, it is easy to understand how this tool will slowly converge to a better and smarter translation tool.

Many CAT tools already allow translators to store their human translations in a database called “translation memory” that can be later prove useful with future translations. However, what is interesting in Google’s technology is that words, phrases or sentences won’t have to be an exact match in order for the system to provide a suggestion: a percentage will serve as a rough indication for how similar the two phrases are alike, which should help the translator’s work.

When translating Wikipedia articles and Google Knols, the translations will be stored in a global, shared translation memory that will be available for everyone to see by default. That means that previously translated phrases from these articles will be stored and will be available for use by other translators, which should allow for a considerably quicker translation process.

Thanks to an agreement reached between Google and Wikipedia, users can download a full Wikipedia entry into Google’s toolkit, translate it into a local language using automated translation, make the necessary corrections and then post them to the localized version of Wikipedia.

Benefits for Google will mainly be in the improved quality of the online translation of blogs and other websites across a number of different language — the toolkit already supports nearly 50 of them.