I'm not exactly an expert on XML processing, however I did some reading myself, on seeing your question. Here's a post XML::Twig tag conversion similar to yours which has been answered by Michel Rodriguez himself (author of the XML::Twig module). The approach given here would be of help to your question as well.

I am not sure what you want to do exactly. If you could provide a better text case, that is code I can run,, with input and expected output, then I would have a better chance at giving you an answer. It would essentially convert your explanations into code, so there is much less chance for misinterpretation. I suspect the solution involves either subs_text or mark but I can't say for sure.

A example test would be:

#!/usr/bin/perl 

use strict;
use warnings;

use Test::More tests => 1;
use XML::Twig;

my $doc='<paragraph> Some <bold>text</bold> here which may be any <bol
+d>length</bold> and <bold>contain</bold> a number of child tags.</par
+agraph>';
my $expected= '<paragraph> Some <bold>text</bold> here which <bold>may
+</bold> be any <bold>length</bold> and <bold>contain</bold> a number 
+of <bold>child</bold> tags.</paragraph>';

my $t=XML::Twig->new->parse( $doc);

$t->root->mark( qr/(may|child)/, 'bold');

is( $t->sprint, $expected, 'simple replace');
[download]

Thanks for replies.

So I'm attempting to alter the string text of <paragraph> without affecting the context of the <bold> tags. Right now, the '->text' command retrieves the text of all sub-elements. Hence, when I set the <paragraph> text, the code is replacing the <bold> elements with a string. I lose the <bold> tag.

I need to maintain the integrity of the <bold> tags while making string substitutions within the <paragraph> element. I will not always know how the text and text children of the <paragraph> element will look. So the content will vary.

I need to be able to find all of the text children of a mixed content <paragraph> element and substitute within the text string without disturbing other child elements.

Example:

Input:

<paragraph> Some <bold>text</bold> here which may be any <bold>length<
+/bold> and <bold>contain</bold> a number of child tags.</paragraph>

Expected output:

<paragraph> Some <bold>text</bold> here which may be any <bold>length<
+/bold> and <bold>contain</bold> a quantity of child tags.</paragraph>

Code:

use XML::Twig;

my $file = '<paragraph> Some <bold>text</bold> here which may be any <
+bold>length</bold> and <bold>contain</bold> a number of child tags.</
+paragraph>';
my $twig = new XML::Twig(TwigHandlers => {'paragraph'  => \&paragraph}
+,TwigRoots => {paragraph => 1});

$twig->parse($file);

$twig->print;

sub paragraph {
    my ($twig, $para,) = @_;
    my $para_text = $para->text;
    &choiceReplace($para,$para_text);
                  }

sub choiceReplace {
     my ($para,$para_text) = @_;
        my $search = "number"; #setting search and replace for example
        my $replace = "quantity";
#locate each occurrence of search term and prompt user to replace
    foreach ($para_text =~ /$search/) {
       
     my $new_version = $para_text;
     my $offset = 0;
     my $new_offset = 0;
         my $result = index($para_text, $search);
     my $new_result = index($new_version, $search);
     $offset = $result;
     $new_offset = $new_result;
     my $l = length ($search);
     
         #loop through string search results while found
     while (($result != -1) && ($new_result != -1)) {
                #create visuals for user to accept/deny match
        print "\n\nCurrent Version:\n $para_text";
        my $replace_match = "**[[$replace]]**";
        substr($new_version,$new_offset,$l) = $replace_match;
        print "\n\nMatched Version:\n $new_version";
        print "\nWould you like to make this change (y or n)? ";
        chomp($change=<STDIN>);
        
        if ($change eq "n"){  
              my $nm_result = rindex($new_version, $replace_match);
              my $nm_l = length ($replace_match);
              my $no_match = "[DENIED]";
              substr($new_version,$nm_result,$nm_l) = $no_match;
                           }
        
        elsif ($change eq "y") {
              my $nm_result = rindex($new_version, $replace_match);
              my $nm_l = length ($replace_match);
              my $no_match = "[CHANGED]";
              substr($new_version,$nm_result,$nm_l) = $no_match;
              substr($para_text,$offset,$l) = $replace;
                    }
#update search starting point
        $result = index($para_text, $search, $offset + 1);
          $offset = $result;
        $new_result = index($new_version, $search, $new_offset);
          $new_offset = $new_result;
                                    }
#set text for <paragraph>
        @_[0]->set_text($para_text);  
                        }
            
                }
[download]

The fact that you want the user input for each substitution makes the problem a _lot_ more tricky. Otherwise you could simply use subs_text as in my previous answer.

In this case you should do the substitution on the text of the '#TEXT' (or '#PCDATA') children of the paragraph. But then what happens if the text contains twice the string you want to replace, and you want only to replace the second one? The logic becomes quite a bit more complex. Of the top of my head I would modify the regexp, to let it skip the appropriate number of occurrences of the string to replace.

A non interactive version that you could use as a basis would be:

#!/usr/bin/perl 

use strict;
use warnings;

use Test::More tests => 1;
use XML::Twig;

my $doc = '<paragraph> Some <bold>text</bold> here which may be any <b
+old>length</bold> and <bold>contain</bold> a number of child tags.</p
+aragraph>';
my $exp = '<paragraph> Some <bold>text</bold> here which may be any <b
+old>length</bold> and <bold>contain</bold> a quantity of child tags.<
+/paragraph>';

# I got a little fancy here to allow several keywords to replace
# the keywords are grouped in a regexp, sorted by inverse length so th
+e alternation works properly 
my $replace = { number => 'quantity' }; 
my $keywords= join( '|', map { "\Q$_\E" } sort { length$b <=> length $
+a } keys %$replace);

my $t=XML::Twig->new( twig_roots => { paragraph => \&subs_word })->par
+se( $doc);

is( $t->sprint, $exp, 'one change') ;


exit;

sub subs_word
  { my( $t, $para)= @_;
    foreach my $text_elt ($para->children( '#TEXT'))
      { my $text= $text_elt->text;
        if( $text_elt->text=~ m{\b($keywords)\b})
          { $text=~ s{\b($keywords)\b}{$replace->{$1}}g;
            $text_elt->set_text( $text); 
          }
      }
[download]