13
votes
Translation services
Does anyone have any idea on how different online translation "services" actually rank now? I was thinking about this today (after I saw the TranslateGemma announcement) and realized that I had not really updated my view on translation apps/services in quite a while.
There is Google Translate, Apple Translate, Kagi Translate, DeepL, etc., but I have no idea how these would rank, especially if it comes to different use-cases.
DeepL is definitely straightforwardly better than Google Translate at the European languages that are its mainstays (it's a German company and I've been using it for German since back when it didn't have many other languages). Chinese and Japanese seem to have more parity between the two (in that they're both somewhat worse on both services).
I haven't used Kagi translate extensively enough yet to have an impression of it, but it seems decent so far. I do like the explicit settings it has regarding the gender of both parties and language level of the output in addition to the formality, as well as the bit where you can add context to the prompt. These seem like very useful settings to me, but I haven't experimented with them yet.
Can't speak much about overall ranking but I do know that the Naver Papago app tends to work better for English <-> Korean translations. Since so much of Korean speaking relies on how respectful you want/need to be to the person you're speaking with, Google Translate tends to fall flat here. Papago lets you pick and choose so it tends to work better in my opinion.
From my experience (mostly translating German and a bit of Japanese) Kagi Translate is the best followed closely by DeepL. Google Translate has improved a lot but is still a ways behind. Kagi has a lot more options for fine tuning the output and sounds more natural than DeepL but is also a bit more buggy.
What do you mean by buggy? Problems more with the site, or with the translations?
Occasionally it will not translate at all and just spit out the input exactly. Kagi also lets you choose between synonyms for individual words in a translation but this feature sometimes also breaks in a similar manner.
These are relatively rare and I still prefer Kagi over DeepL. I think DeepL just benefits here from being around longer and having a bigger team.