How to prevent upper case capitalization for certain words at the end of a song title?

So I've got a "capitalization" action consisting of several capitalization rules such as regular expressions to ignore the upper case (applied typically for every first letter in a word) for certain words (in, at, for with etc.).
This covers both english and greek words and seems to work fine. However the rule to upper case the first letter in the last word no matter what (even for articles and prepositions for example) is not valid in greek (upper case for the first word still stands and same goes for the last word IF it's not an article/ short preposition etc.).
How can I make sure the action follows a regular expression such as:
( στις| της|)\b
$lower($1)
for the words included even if they are located at the end of the title?

If you have a list of words that have to be lower case, it would be easiest to create an action group of simple replace actions in which you can set that only words should be treated and that the case should not be ignored.

To get the last string in lower case, you could try:
$regexp(%title%,'(.*) (.*?)$',$1 \l$2)

Thanks for the reply. After several attempts of trying to figure this out on my own, I have to admit defeat and ask if you could please explain in more detail. I don't really understand working with regular expressions etc., I merely use an action group I've found in here years ago (with a few adaptations to my liking).
So, first of all should I ditch my action group completely and start a new one?
Creating a new regular expression action like that
seems to do nothing other than delete (replace with an empty space I'd assume?) the word at the end of the title I wanted to be lower case.

The action type would be "Format value" for the field.
I would run this action as the last one in the case conversion action group as it does nothing but replace the first letter of the last word with a lower case one.
If there are, however, just a few words that could be at the end of a field and that should always be in lower case, I would create an action group with one action per word that replaces the upper case word with the lower case word - so this is not "case conversion" but "replace".