Find Unclosed Quotes (‘- or “-Style) - regex

Question

I'm editing some text directly from OCR engine and in some paragraphs the OCR engine ignores the opening and closing quotes. I prefer editing in HTML mode and as a result end up with some text like:
<p>“Wait a moment,” Jacey said. The street light lit up his aged, rat face. Who’s on the move?”</p>
Notice the missing “.
Another sentence:
<p>“He said he’ coming afer you,” Harry said, and he’ bringing the boys too!”</p>
I use this regex : ([>\.\,])(.*?)” which seems to do the job for the second sentence but not for the first. This is because the regex is matching from left to right and so matched the extra sentence The street light lit up his aged, rat face. which should not be within the quotes.
I was thinking that the problem can be solved if the matching was done from right to left. I know this is an option available in C# but I'm using the regex engine of text-based editors to edit a simple text file. Is there a way to locate just the last sentence before the “, which is the sentence Who’s on the move?.
[EDIT]
I have been trying using the lookbehind regex: (?<=(?:\. |, |>)(\w)(.*?))(”) which seems to find all sentences with missing open quotes, “, but the problem is I cannot replace the contents inside the (?<=) construct with \3“\1\2\3 because lookbehind is 0 length. Instead the text is just duplicated. For example with the above regex the sentence Who’s on the move?” becomes Who’s on the move?”“Who’s on the move?”
Any ideas will be appreciated.
Thanks

Answer 1

Recursion and Defined Subroutines
The following regex checks that strings are balanced. The code below (see its output in the online demo) checks several strings. The explanations are in the comments.
$balanced_string_regex = "~(?sx) # Free-Spacing
(?(DEFINE) # Define a few subroutines
(?<double>“(?:(?!&[lr]squo;).)*”) # full set of doubles (no quotes inside)
(?<single>‘(?:(?!&[lr]dquo;).)*’) # full set of singles (no quotes inside)
(?<notquotes>(?:(?!&[lr][sd]quo;).)*) # chars that are not quotes
) # end DEFINE
^ # Start of string
(?: # Start non-capture group
(?&notquotes) # Any non-quote chars
&l(?<type>[sd])quo; # Opening quote, capture single or double type
# any full singles, doubles, not quotes or recursion
(?:(?&single)|(?&double)|(?&notquotes)|(?R))*
&r\k<type>quo; # Closing quote of the correct type
(?&notquotes) #
)++ # Repeat non-capture group
$ # End of string
~";
$string = "“He said ” ‘He said ’";
check_string($string);
$string = "<p>“Wait a moment,” Jacey said. The street light lit up his aged, rat face. Who’s on the move?”</p>";
check_string($string);
$string = "<p>“Wait a moment,” Jacey said. The street light lit up his aged, rat face. ‘Whos on the “move?” ’</p>";
check_string($string);
$string = "<p>“He said he’ coming afer you,” Harry said, and he’ bringing the boys too!”</p>";
check_string($string);
$string = "<p>“He ‘said he’ coming afer you,” Harry said, and he“ bringing the boys too!”</p>";
check_string($string);
function check_string($string) {
global $balanced_string_regex;
echo (preg_match($balanced_string_regex, $string)) ?
"Balanced!\n" :
" Nah... Not Balanced.\n" ;
}
Output
Balanced!
Nah... Not Balanced.
Balanced!
Nah... Not Balanced.
Balanced!
Replacing Missing Quotes
As I've indicated in the comments, IMO replacing missing quotes is hazardous: before or after what word should the missing quote fall? If there was any kind of nesting, can we be sure that we've correctly identified the missing quote? For that reason, if you're going to do anything, my inclination would be to match the balanced portion (hoping it is correct) and remove any extra quotes.
The pattern above lends itself to all kinds of tweaks. For instance, on this regex demo, we match and replace an unbalanced quote. Since this was requested, I'll offer a second potential tweak with some reluctance—this one inserts a missing left quote at the beginning of the phrase preceding the unmatched right quote.