imperfect translations corrected by another

Oh that would be a pain to develop this. It means we would ask to the experts to rate the quality of the target language and to rate the understanding of the source language. That would make the process more heavy (and very heavy to develop too) but it can work. Anyway, ratings are not the most important thing on cucumis. I'm not sure we should spend a lot of effort on this.

I mainly focused on target laguages when I designed cucumis because of what you explained Kafetzou. It's possible to understand a text even if you are not fluent speaker. But it's not possible to write correctly in a language you are not fluent. That's why the translator is asked (warning displayed before the translation) to be a fuent speaker of the target language. I'm not sure it's necessary to ask the same thing for the source language.

It depends on what you need the translation for. If someone receives a message in a foreign language and doesn't understand it (or he is not sure to understand idioms, acronims, slang) it should be better for him to have it translated by someone who fully understands it, no matter if it doesn't put it perfecty in the target language, because in this case the requester wants to be sure to understand all the text.
On the contrary, if one needs a translation of a letter he has to send, he should want it to be written by a native speaker of the target language.
(By the way, I made more translations into English, than into Italian...)

Actually, I don't think it would be that hard to develop. You could simply take the points earned for a translation and drop them into two calculators - one for the target language and one for the source language. The rating system could remain the same, but there would be another number that shows up for each translator, that would be source language ratings.

Qu'est-ce que c'est "drop"? Désolé, mais je n'ai pas bien compris ton dernier message...

C'était primairement pour jp, mais je vais essayer de le traduire.

Je crois que ce ne serait pas trop dur de développer. On pourrait prendre le numéro de chaque traducion et le diriger (ou laisser tomber - ça c'est "drop" ) à deux calculateurs - l'un pour la calculation de l'évaluation de ce traducteur en la langue de départ et l'autre pour sa évaluation en la langue d'arrivée. Et pour chaque langue qu'il traduise, chaque traducteur tendrait deux numéros - l'un pour la langue de départ et l'autre pour la langue d'arrivée.

Merci beaucoup kafetzou! Maintenant je comprend mieux... Mais je ne suis pas encore sûr d'avoir compris (à niveau pratiquq) au 100% le point "calculation de l'évaluation de ce traducteur en la langue de départ". Par exemple, je traduis un text français vers l'italien: moi j'aurai un "rating" en italien et en français, CAD, ma comprehénsion du français et ma capacité de parler/écrire l'italien?

I will keep this suggestion in my todo list as it makes sense. But it's medium priority for now as I don't think the ratings are so important.

Et maintenant c'est moi qui n'a pas compris. Qu'est-ce que c'est "CAD"?

C'est À Dire

A - merci! Le "rating" serait le même numéro, car les expertes ne peuvent pas donner deux ratings pour chaque traduction. Par exemple, l'experte de l'italien te donne un rating de 8 - ce numéro apparaîtrait en deux endroits - pour l'italien comme la langue d'arrivée et pour le français comme la langue de départ.

I think this will not work: one may write a perfect Italian, but not related at all to the French source text.

What really shows that you can understand a language is that your translations are accepted.
On the other hand, the fact that a translation is rejected may be not due to the translator's bad skills to understand the original meaning. It's not so easy-going...that's my opinion.

This is exactly my point. For example, you may have had a lot of translations from Hindi into Portuguese accepted, implying that you understand Hindi pretty well, but I would never know that, because I can only see your points for Portuguese.

This is important to me as an expert, because when I ask for help evaluating a translation, and I don't know the people who've expressed an opinion on it, I usually look at their point values to see what languages they seem to know.

This just happened to me a few minutes ago. I'm trying to evaluate a translation from Hindi into English, and the English looks fine, but I can't even see the Hindi (it doesn't come up on my computer), and even if I could, I couldn't read it, so I asked for help.

Someone whose name I'd never seen before expressed an opinion on the translation, and when I looked at her profile, I saw that she claimed Hindi as both a source language and a target language, but she has no points for Hindi, so how do I know if she really does know Hindi?

With the system I propose, I would know.

That's going to increase the workload on the experts to rate on both the source and target language.

We already have the system for rating the source language, and that's "Translation Accepted"(Understood the source), and "Translation rejected" (Is totally clueless). We only lack graphical presentation showing how many translations done were rejected and how many were accepted for each member.

kafetzou, you can get into the person's profile page and take a look at the translations she has done. If she has done translations from Hindi, we can safely infer that she understands Hindi.

Oh - thank you, Samantha - I hadn't thought of that! I'm still stuck, though - she's done two from Hindi, but they were both from Hindi in Latin script.

As for increasing the workload, I think you didn't read my posts carefully regarding my idea. I didn't say that the experts should give two numbers - I said that the one number they give should go into two calculators - one for the source language and one for the target language.

But you're right - we don't need that as urgently as I had thought because there is a way to find out people's competency in the source language(s).

For now the rating is reflecting only the target language. That mean you can reject a beautiful english translation with a 10 because the meaning is false.

When you accept a transaltion, that means the translator has well understood the meaning of the source text but that also maybe mean you have edited it to make the meaning right.

Well you see it's not easy to take conclusions about the source language if the experts don't give a rating about it.

A solution can be an automatic percentage about accepted translations for every source lang.

CocoT a déjà écrit un message à ce propos, Xini. ça devrait être dans un forum anglais... I'm searching... Here

Ooh, maybe a lot of people already thought about something like this...