Hacker News Reader

Comment by mdasen | Hacker News Reader

https://www.strakertranslations.com/translation-pricing/ - $0.13 - $0.16 per word which comes to $32.50 - $40 per 250-word page.

https://gengo.com/pricing-languages/ - $0.06 - $0.12 per word or $15-30 per page.

https://www.gts-translation.com/translation-prices-per-word/ - $0.10 per word or $25 per page.

https://www.onehourtranslation.com/translation/benefits/tran... - $0.087 or $21.75.

It's certainly seems cheaper and Bitcoin fees seem relatively low right now (at least compared to the end of last year). However, I'm not sure it seems so significantly cheaper, especially weighing the 250-word one-size pricing.

Agreed that it's really cool, though.

View on HN

WOW... no wonder there are so many AI startups working on translation.

I thought Google Translate uses AI now fairly perfectly.

Why is anyone paying a human translator $40/page when Google Translate is "free"?

There are plenty of situations when the human sense of nuance and appropriateness is absolutely necessary. Translation of legal documents is at the extreme end of this (for example, when my Brazilian wife had to translate her university diploma into English upon moving to Sweden.)

If you just need to get the basic gist of something in a different language, then machine translation is obviously more performant.

Because even Google and Microsoft's translation service isn't really that good, if you want an idiomatic and context-sensitive translation. I'm only involved in some efforts trying to use these services for real-time text translation for IT helpdesk solutions, where results are... mixed, but close enough to usable, depending on the language, but I can easily imagine that for other domains you would definitely want a real person who could evaluate the nuances.

Google Translate still gets a lot wrong. It's good enough to understand if you just need to translate foreign language text to something you understand. But for professional documents it's nowhere near adequate.

Another point is it misses a lot of context. My girlfriend is fluent in English and Chinese and has an Art History degree. That domain knowledge makes her quite valuable as a translator for Chinese galleries. There's quite a lot of research involved when she does a translation to find the correct terms for techniques, mediums, styles. Translation is more difficult than it would seem on the surface.

Google Translate is good for like, high school language elective level translations. For a personal use (e.g. buying from a Spanish website) this is more than sufficient. However it is still very easy to tell machine translated text from professionally done work.

Google Translate still screws up and can frequently sound a little "off" to a native speaker. Professional documents are still going to need a real translator to give them a once over at a minimum.

Google Translate covers the case that previously would have gone untranslated: e.g. I'm interested in reading this article, but it's in Italian, oh well, I guess I can't. If you're actually publishing something, you need a human, and it's going to take them time and effort and cost you money.

BTW, I find for German-English, DeepL is better than Google or Bing. That may be the case for the other languages it supports, too. Still not good enough for anything that needs to be correct.

Google translates "The IT group" to French as "Le groupe IL". Context matters.

That's good to know!

I haven't really used them for anything serious in the past few years which is why I asked.

I had just read all of this AI stuff recently about Google using machine learning/AI to better help it's instant audio translation service or something.

Perhaps the only thing they machine-learned was the audio to text conversion part? Not the translation part?

Reply on news.ycombinator.com