Two more AI sites Uwekaji wa uzi: Samuel Murray
|
|---|
Samuel Murray 
Uholanzi
Local time: 04:09
Mwanachama(2006)
Kiingereza hadi Kiafrikana
+ ...
|
And?
I had two emails from young ladies of ill repute in my mailbox today but I’m not sure how that will help other translators, beyond a little stress relief perhaps. A bit like AI then. Entertainment value.
[Edited at 2023-04-25 15:33 GMT]
| | | | Samuel Murray
Uholanzi
Local time: 04:09
Mwanachama(2006)
Kiingereza hadi Kiafrikana
+ ...
KIANZISHI MADA
Ice Scream wrote:
I had two emails from young ladies of ill repute in my mailbox today...
Tsk-tsk... you forgot to post the URLs.
| | | | Philip Lees
Ugiriki
Local time: 05:09
Kigriki hadi Kiingereza
Samuel Murray wrote:
A tool that uses AI to evaluate the differences in MT output:
https://www.machinetranslation.com
Near the top it tells you what the differences between MT outputs are (this is not AI driven), and then at the bottom it asks Chat GPT to comment on the accuracy of each translation.
I tried it with a piece of text from a Greek newspaper. This is the translation delivered by ChatGPT, which my eyeball inspection identified as the best, for reasons I won't go into unless asked.
These are the places that are most at risk from heatwaves. The dangerous heatwaves - which are breaking records - are set to increase as the climate crisis intensifies and will be particularly devastating in countries that are less prepared, according to a new study. Scientists analyzed temperature datasets covering more than 60 years, as well as climate models, to calculate the likelihood of unprecedented extreme temperatures - and where this could occur. They identified Afghanistan, Papua New Guinea, and Central America - including Guatemala, Honduras, and Nicaragua - as the "hot spots" for high heatwave risk.
Pretty good. A faithful translation of the original in terms of both content and tone, with only one minor error. The first sentence was the headline for the article, but ChatGPT was the only one that treated it correctly. The other engines either ignored it altogether (ModernMT), or stuck it at the beginning, running on to the next sentence without a space or any punctuation. This may have been an artefact of the website and the way it handles the text input, rather than an intrinsic fault of the translation software, although I made sure there was a blank line in the source between the headline and the rest of the text.
The evaluation software identifies the headline, "These are the places most at risk from heat waves", as an "identical" section, even though three of the engines return "heat waves" as two words, whereas the others have it as one (ModernMT completely ignored both the headline and the final sentence, for some reason.) The "AI translation analysis" at the end was left blank for my example.
All very well, but this is the kind of text I would expect a large language model to do well on, as there is plenty of comparable material out there. Furthermore, as I've said before in other threads, MT is fine ... until it isn't.
| | |
|
|
|
Tom in London
Uingereza
Local time: 03:09
Mwanachama(2008)
Kitaliano hadi Kiingereza
Samuel Murray wrote:
A tool that uses AI to evaluate the differences in MT output:
https://www.machinetranslation.com
Near the top it tells you what the differences between MT outputs are (this is not AI driven), and then at the bottom it asks Chat GPT to comment on the accuracy of each translation.
I tried that one using an extract from a document I'm working on (Italian to English): a badly written technical and user manual for a piece of machinery (an industrial-scale water purification plant).
As often happens, my job in translating it is to make it more comprehensible in English than it is in Italian.
Chat GPT doesn't know whether the translations (in Deepl and Google Translate) are accurate or not because it doesn't understand that the original document is badly written.
[Edited at 2023-04-26 08:27 GMT]
| | | | Baran Keki
Uturuki
Local time: 05:09
Mwanachama
Kiingereza hadi Kituruki
| | | Tom in London
Uingereza
Local time: 03:09
Mwanachama(2008)
Kitaliano hadi Kiingereza
| Two more AI sites | Apr 28, 2023 |
Is it possible to set up two or more AI sites so that they chat to each other, and we don't have to bother?
| | |
|
|
|
Samuel Murray wrote:
A tool that uses AI to evaluate the differences in MT output:
https://www.machinetranslation.com
Near the top it tells you what the differences between MT outputs are (this is not AI driven), and then at the bottom it asks Chat GPT to comment on the accuracy of each translation.
This one is actually pretty interesting, though ChatGPT's comments aren't useful at all.
| | | | To report site rules violations or get help, contact a site moderator: You can also contact site staff by submitting a support request » Two more AI sites | LinguaCore | AI Translation at Your Fingertips
The underlying LLM technology of LinguaCore offers AI translations of unprecedented quality. Quick and simple. Add a human linguistic review at the end for expert-level quality at a fraction of the cost and time.
More info » |
| | TM-Town | Manage your TMs and Terms ... and boost your translation business
Are you ready for something fresh in the industry? TM-Town is a unique new site for you -- the freelance translator -- to store, manage and share translation memories (TMs) and glossaries...and potentially meet new clients on the basis of your prior work.
More info » |
|
| | | | X Sign in to your ProZ.com account... | | | | | |
Login to reply/comment 
