Open dgilge opened 6 years ago
Hi Daniel, In the case where we have an unofficial translation, we should use it as the basis, as it is most probably better than google or deepl at this point. I think we should only have one button (option) for machine translation. We should just choose the best one for them. Which would be deepl I guess.
Hey @Salomito, make sure to remove your email signature for GitHub replies :)
I agree, we only offer one machine translation option besides prior translations. I need to think about a good way to integrate this into the UI
Thanks for you comments. I just deleted the e-mail signature (also in the history).
Alright, I agree that simpler is probalbly better.
Do I understand you correctly to still be able to choose between an old translation and a machine translation?
In this case I'd create a separate database table for that. But I think I'll do that anyway in order to have a vocabulary for autocompletion (and spellchecking).
I am not sure if it would be advisable to let the person choose between old translations.
It is better to ask persons of experience which old translation to offer, or to have some way of letting the users choose beforehand.
Let's discuss the details another time.
I'll use a separate table here. Not last because of statistics. (When I copied the machine/old translation directly into the translation segments the statistics said that everything was translated.)
Do you have a good name for that? This is what I need right now. I'm thinking of "TranslationResource".
"theOldAndTheUgly"
This is not specific enough. It could be everything. Well, as long as it isn't new and pretty. ;-)
BaseTranslation
There's something else to consider: Identical sentences/paragraphs which are translated in another book already
Oh yes, that's right. Especially with all those compilations. Definitively important for EGW books.
We decided to have an additional API and maybe an extra table (BaseTranslation
, in case this is more practical than direct access to the translator API) for DeepL, etc.
I'll come back when start implementing this and I need API access for DeepL. The API can be found at https://www.deepl.com/api.html
Issue added to July milestone today.
We also said to have a button in the UI to select the AI translation as basis for a human translation.
One thing I thought about is to be able to read the AI translation. The reason for this is simple: It might be that somebody wants to read some text in his language which is available as AI translation, but he had to wait several years for a human translation.
What do you think? Can we make a UI for AI translations (as long as there is no human translation)?
My suggestion is to let the user choose what (and if) he wants to use as basis for his translation:
He can set a default in his (a) user settings and for (b) every book. But he can still see all availabe translations for (c) each paragraph.
This is a priority list for (a) and (b): In case the first is not availabe the second (or
n
th) will be used.What do you think? As soon as I have your responses I can go on…