Lilt: faster translation than SDL Trados

In a newly-released comparison of computer-aided translation tools, initial research found human translators improved their translation speed using Lilt on average between 13% to nearly 22% over SDL Trados while producing comparable quality output. The translation staff here at e2f played a central role in the study.

The pilot study comprised a proctored environment focused on real-world production conditions for machine translation postediting. It monitored e2f’s team of English-to-French translators at our San Jose offices. Our translation team worked on two datasets to translate: user interface (UI) help files and hotel chain content.

Results in terms of times saved for both were significantly better for Lilt over Trados, while translation quality did not suffer. The fastest translation speed measured using Lilt was a raw 1,347 words per hour, which was a stunning 33% improvement over the fastest comparable work produced using Trados.

One of the e2f translation team members shared verbatim insights in a post-study interview. “The engine improved as we translated. The suggestions got better and better as we used it.”

In terms of accuracy, the translator was satisfied. “It is very obvious when you copy an entire sentence that was translated correctly. You increase your productivity by ten, fifteen seconds or more per segment.” In the end, it was the combination of each of these line-by-line speed improvements that resulted in the large percentage gain over SDL Trados.

Read the entire study and its results, available online at Lilt.