Coooooool, Mastodon traduit directement les textes en alphabet grecs des compas que je suis.

Bon, c'est traduit par Deepl.com mais au moins c'est pas Google, non ?

J'avoue ne rien y connaitre, mais je me dis que si masto l'utilise, c'est que c'est moins pire que pire...ou bien ?

#Deepl seems to be working rather well

DeepL integriert seine Voice-Übersetzung in Zoom – und erweitert um Mandarin, Ukrainisch und Rumänisch. Wird der Kölner Anbieter damit zum neuen Standard für Live-Kommunikation? Ein spannender Schritt für globale Zusammenarbeit. Mehr dazu hier: #DeepL #Zoom #KI
https://www.all-ai.de/news/news24/deepl-voice-mandarin-zoom

I've now apparently got translation services on my server via #DeepL

We'll see how it goes

DeepL Übersetzungen werden immer beschissener. Die KI baut ungefragt ganze Sätze um, zum Teil absolut Sinnfremd, und welche nichts mehr mit dem Ausgangstext zu tun haben. Anfangs war das Tool hilfreich zum Erlernen neuer Sprachen, aber inzwischen entwickelt es sich zu einem Albtraum. Es zeigt sich immer öfters, dass KI mehr Probleme schafft, als sie welche beseitigt.

While #Fedilab embeds its own translation feature (default option uses our #LibreTranslate service), you can also use #Lingva or #DeepL if you have an API key.
But we have to add the ability to let you use your instance as an option to translate messages. That will be done.

Desvende o Poder da Modelagem com AutoKeras e Keras Tuner!
Quer simplificar sua jornada no deep learning? Neste post, exploramos como o AutoKeras e o Keras Tuner podem transformar a modelagem de suas redes neurais, facilitando desde a seleção de hiperparâmetros até a criação de modelos eficientes. Não perca a chance de alavancar seu projeto! Clique e descubra mais!
.
.
.
#AI #DeepL...
https://inkdesign.com.br/autokeras-e-keras-tuner-simplificam-modelagem-em-deep-learning/?fsp_sid=90856

@peertube There seems to be a change in the #weblate update process of translation strings (now twice in a row). Unfortunately this is to the worse. The English strings are copied to the target language strings automatically which makes it much harder and works against weblates support of your workflow in many cases. Please keep the *new* target language strings empty like previously.
The #deepl "too many glossaries" error seems to be partially fixed though. Cheers.

#WritersCoffeeClub July 8: Have you ever received feedback on a piece that completely missed the mark?

The only thing that comes to mind is: "Have you ever considered using #AI -based #translation for your work?"

While yes, my writing work is all about translating all German folk tales, the actual "translation" aspect is the least part of it. And even here I don't see AI-based translation as particularly helpful, since the 19th century texts I translate are full of obscure idioms and dialect words that AI translation would struggle with.

But more importantly, when I translate these texts I have to think about the _context_ of these tales - the 19th century environment in which they were set and told. And then I have to research and explain this context to an international audience.

Can AI do _this_ part for me? Without constantly hallucinating blatant falsehoods? I have serious doubts.

No, if I want to do all this _right_, then I have to do this work by myself. #amwriting #DeepL

I'm fully getting mad when interacting with one of the #PocketID developer, he use machine translator provided by #Crowdin and #DeepL to filled up all of the missing strings of the translation, the translated strings are really really bad, so I submitted an issues to asking them to stop this, but he "Closed with no planned".

OK so LibreTranslate is part of the mastodon instance I'm on.

I just saw a post in Chinese:

在贵州被虫子咬了眼皮，上次这么惨还是幼儿园。

LibreTranslate gave this:
In your State, the puppet was pested, and last was a kindergarten.

DeepL gave this:
I got bitten by an insect on my eyelid in Guizhou. The last time I had such a bad experience was in kindergarten.

I suspect that of the two, DeepL is the more accurte haha

Juhuu! Endlich hat mas.to DeepL @trumpet
#DeepL

Wie wirken sich KI-Übersetzungsprogramme auf die Forschung in den Geisteswissenschaften aus?

Eckhart Arnold fasst zentrale Erkenntnisse aus einem deutsch-tschechischem Workshop zusammen: Zum Beispiel, dass die Nationalsprachen als Wissenschaftssprachen wieder gestärkt werden könnten, wenn Forschende in der Muttersprache schreiben und dann übersetzen lassen

https://dhmuc.hypotheses.org/3493

When translating into certain languages, #DeepL offers a switch for ‘formal’ or ‘informal’ language. I assumed that this means that pronouns like tu/du/jij etc. will be used in informal mode, whereas vous/Sie/u etc. will be used in formal mode.

Recently I encountered an ‘informal’ #translation of ‘Please complete all content above before continuing’ into German: ‘Mach erst mal das da oben, bevor du weitermachst.’ (Roughly: ‘Do the stuff above before moving on’)

What’s going on here?

#GoogleTranslate seems to be getting worse and worse, sometimes omitting entire sentences or ignoring meaningful words.

Hilariously, it also attempts to translate all-digits dates but then messes them up so they become a completely different date somehow. E.g. 2025-05-26 somehow becomes January 1996

#DeepL is going the other way, getting quite good now, even at more obscure stuff like blog posts with wordplay and spelling errors.

This observation is based mostly on Lithuanian/English examples.