Navigating the Web of Translation with Linguee

How many times have you tried to translate a sentence and have come up with embarrassingly incorrect results via a translation site? Or have you used Google Translate (s goog) to translate a website from another language and cringed at the stilted results? Most of us who need translations for our work struggle with the available tools out there or bite the bullet and pay someone to do the translation work.

Linguee is a new web-based translation tool on the that combines artificial intelligence with human intelligence to bring more accurate, relevant and nuanced translations to bear. Linguee sifts through hundreds of millions of translations aggregated from high-level multilingual sources on the web, such as documents from the European Union and Canadian governments. The site’s “machine learning system” sorts and presents results for the variety of possible translations and users of the system can then rate definitions with a thumbs-up and thumbs-down system. Those human ratings “teach” the system to rate not only the particular entry but to rate the source as well.

The site is especially good at handling words with many meanings and gives you search results that include an expandable list of actual sample uses of words in context from a variety of sources. Next to each word translation is a pie chart showing the percentage of usage of each translation to identify the most common uses of the translated word.

Linguee also handles searches for common phrases and provides you with validation that your own translation is correct or gives you additional options based on the most common translations the system finds. The site is being used by professors, professional translators and others whose jobs depend on the accuracy and appropriateness of translations. Even for someone who has a working knowledge of another language, Linguee can add some extra assurance that one’s translations are correct.

Linguee’s tagline is “The web as a dictionary;” it melds machine processing and algorithms with crowdsourcing human input to deliver better translation results.

How are you translating content, and what kinds of materials are you needing to translate?

Related GigaOM Pro content (sub. req.): Report: The Real-Time Enterprise