Don't translate Scripture references in non-verse text

[johnml1135]

There are often Scripture references at the beginning or end of non-verse text. Currently we are trying to translate these and making a mess of things. We should be able to do better. Here are some options:

Best proposal:

• Strip out all references when making training/pretranslating data
• Re-insert them when putting into USFM

Complications:

• How do we know something is a scripture reference? Can we truly make a regex to capture all references? What about multiple languages?
  • This one should work for english: https://stackoverflow.com/questions/66818564/regex-for-bible-references
• Re-insertions would require reworking the USFM updater - which may add some complication (but not too much because there will be no ranges do deal with in non-verse scripture text).

[ddaspit]

We should just be more careful about what markers we translate and what markers we don't translate.

[johnml1135]

Yes - we could try some funny things with regex's - but I think just the "translate these tags" and "don't translate these tags" is the best way to get to the 95%.