Yulia Kovalskaya 

I guess you can do this:

• If you can’t get the original document used to create the PDF, you’ll need to extract and clean the text (deleting extra returns). Manual work, I know. Maybe an AI can help you deleting extra returns.
• If the content is big, ask an AI to extract only English text (save it to a new document) and extract only French text (save it to another document).
  • Clean and confirm both monolingual documents (since we agree that AI work well until they fail, don’t we?). Yeah, more manual work here.
• If the content is small, create manually an English-French table (the Italic formatting may help you to move the text to the English column), and save each of the 2 columns into different documents.
• Finally, align both documents in Trados Studio. You’ll still need to confirm the automatic connections. And save the project to an empty TM, which I believe is what you want, don’t you?
  

Yulia Kovalskaya 

I guess you can do this:

• If you can’t get the original document used to create the PDF, you’ll need to extract and clean the text (deleting extra returns). Manual work, I know. Maybe an AI can help you deleting extra returns.
• If the content is big, ask an AI to extract only English text (save it to a new document) and extract only French text (save it to another document).
  • Clean and confirm both monolingual documents (since we agree that AI work well until they fail, don’t we?). Yeah, more manual work here.
• If the content is small, create manually an English-French table (the Italic formatting may help you to move the text to the English column), and save each of the 2 columns into different documents.
• Finally, align both documents in Trados Studio. You’ll still need to confirm the automatic connections. And save the project to an empty TM, which I believe is what you want, don’t you?
  

Dear Jesus! You are so helpful, as always!!!!

But we have many docs in pdf, no original files (word) and...yes, I thought about Copilot....and it returned something really horrible: a .txt, no format, no pics, no tables of contents :)....

Yulia Kovalskaya 

Images, tables of contents, bullet symbols won’t go to the Translation Memory either. Only inline images would go as a placeholder tag.

And formatting from a PDF can be a real nightmare anyway. In the sample screenshot above there is only one instance of formatting.

I´m afraid I have not explained the task properly above. 

The screenshot you see is bilingual.

We have to leave the French part as it is, untouched and translate texts in English to German.....This is where it becomes so complicated....

Yulia Kovalskaya 

So you need to remove French from a document extracted from a PDF. Converters are never perfect, and the best ones are paid ones. And images will be downgraded to whatever resolution chosen when the PDF was created. Your best option is to get the original source documents.

I've heard that the AI Claude (from the company Anthropic) can now handle Word and even PDF files, but I’ve never tried…

https://www.inc.com/ben-sherry/anthropic-says-this-ai-tool-can-now-create-and-edit-documents/91237278

Even worse, Jesús....leave the French texts there...but , yes, extracting was one of the options...and it failed:(

This is why I tried language recognition in Word, rememeber??

For highlighting the texts in one of the languages and then hiding / showing only the colored ones in TRados with Advanced Filter. No way! Once converted to word (we use Adobe Pro), text languages were not automatically recognized by Word....

Yulia Kovalskaya 

Yulia Kovalskaya said:

(we use Adobe Pro)

Did Adobe Pro gave you the text from your screenshot below? I don’t think so…

Yulia Kovalskaya said:

text languages were not automatically recognized by Word

You can take italic formatting as English, with exceptions, and mark it somehow:

Hmmm....you are right...almost all English texts come in Italic.....But then comes another handicap: 

Look, these "inserts" are just lovely...

And then, not all English texts come in Italic...in one word: a nightmare. With no solution. Only contacting the supplier and asking for the original docs...just as Paul said...