Background of the issue:
I am translating from Polish to English using WPML. My main translation engine is Google, followed by Deeplearn and Bing. I expected WPML to use Google Translate for Polish-English translations.
Since all three support Polish to English translations, I expected only Google Translate to be used.
Symptoms:
WPML seems to translate with Bing instead of Google. The translation quality was poor, and it matched Bing's output, not Google's. I checked it manually. I took the same paragraph and used web version of Bing Translate. And to my surprise the output text was identical. So I took the same Polish text and this time inserted into web version of Google Translate. The output was high quality and different. It looks to me that Bing is used even if Google Translate is the first engine.
Questions:
Can you check on your end which engine is actually used when translating any test from Polish to English? (Use Google for primary, Deeplearn for second, Bing for third)
If it is actually Bing being used anyway, how many credits are charged?
The report is not clear, but it is irrelevant. On my end I selected Google, but the output comes from Bing. So even if the report says Google it does not change anything.
By checking on your end I meant the code, not my account.
And the fact you are advising me to disable the engine "to be safe" means that you are actually unsure. So please make sure insted.
If indeed Bing is used instead of Google, when Google in switched on and set as first, then logically I cannot be sure Bing will not be used even if I switch it off...
What if there is a problem with code and both or all 3 engines are named Google, DeepLearn and Bing, but by someone's mistake all of them send data to Bing?
This is what I ask you to check, not my account, not my website, but your code. The output I got was from Bing instead of Google, so please check how this happened, not advise me how I can setup setting on my websites.
I’m here to help you achieve the results you’re looking for.
The report provides a clear and precise way to identify each translation JOB ID and the corresponding post or content. It’s part of the code as it comes directly from the translation management servers, not from your account. It shows what API was executed for each job while using out code.
Regarding your mention of "The output I got was from Bing instead of Google," could you share more details on how you tested this? For example, did you verify it manually with Bing or Google translation tools?
To clarify what I meant by "to be safe," I was suggesting that, in the future, if Bing might be used (even though it’s unlikely), you can avoid this possibility by removing Bing from the translation settings. And it will be removed, it is certain.
Here’s a bit more on how the code prioritizes translation engines:
- The system only defaults to another engine if the primary one is busy or unavailable or if it doesn’t support the chosen languages.
- In your setup, if Google isn’t available due to issues like too many requests or service downtime, DeepL will be used instead.
- Only if both Google and DeepL encounter issues, which is very rare, would Bing be used.
- If you do not have Bing active at that rare occurrence, then the translation will not go through.
