punctuation breaks translation quality

[commandline-be]

Hey,

Thanks for the great software, it's already been useful. I want to report the below though I assume this is well known.

The translation does appear to suffer from something potentially avoidable. This being any punctuation seems to garble translation, may that be practical use of " ' ( ) { } . - /

Notable is when for example "myword-split" happens this causes the translation to show something like '%translated-word but not punctuate it as in the original. There is a hack of sorts possible by inserting a space after the first " for example. This drops the punctuation marks entirely but does improve translation quality considerably.

Surprisingly this use of (word/word2/word3) is translated (tword/tword2/tword3) consistently.

To my understanding none of these symbols are critical to letter or symbols in any grammar for most if not any language so it would be recommendable to integrate a default ignore-so-copy in the translation.

I assume this is in part also why text inside a table suffers from translation considerably. This can be shown by copying text inside a cell to a translation form and submit for translation. This is of notably higher quality.

I'd also argue to not translate target language since especially English is often target language and often observed in source texts as part of for example technical documents. I can open another issue for this should that be required.

Br,

JL