Algorithm for ranking translations needs to be improved
I love the concept behind Duolingo, and am positively addicted to the site after just two days of use. However, the way translations are ranked needs improvement. Oftentimes the most literal translation gets ranked the highest, even if it doesn't sound natural in English. What's the point of having humans do the translation if the end result is just as rigid and inflexible as a machine translation? There are instances where verb tenses need to be adjusted, definite articles need to be switched to indefinite ones, singular nouns need to be made plural, or in some cases, entire words or phrases need to be replaced with something more idiomatic. In spite of this, sentences translated with such stylistic alterations are too frequently ranked inferior to the rigid, word-for-word literal translations.
True enough. This topic however boils up every two weeks and yet no solution so far.
The way I see ranking is done by humans too. The sooner people start ranking the translation with such alterations as good it should be the chosen one. The fact is people coming here are not fluent and the rating is not being done by professional bi-linguals. We need to learn fast and help improve the system by our input don't you think?
We are on it! we have couple of solutions for this problem and we’re currently thinking about how to implement them. In the meantime, don’t hesitate to use the "suggest edit" feature to let the author of a translation know about your change.
Dear naut, thanks for the heads-up. Please do not take my comment as too harsh a criticism ;) I think we're all looking forward to those improvements. Problem with the "suggest edit" is the shear number of translations. So there is a question I have with this: If such a suggestion is made to a certain users translation is that suggest also applied to all other users translation which are the same? I hope you get what I'm trying to say here ;) Kind regards
@Yahp: Yes, I think a suggest edit is sent to all translators. I had the case where someone suggested an edit on a heading and several people commented afterwards on the same suggestion. That indicates that this suggestion was sent to all the people who came up with that particular translation.
Isn't the original question more a human than an algorithmic problem? After all people rate the literal translation as good and do rate the more free translation as high.