The way I see it, you have two possible approaches.

You could construct an index where each word of interest is stored with the meta information that it is associated with.

open $srcFile, '<', 'source.txt' or die $!
my $metaText = "";

while( my $line = <$srcFile> )
{
    if( $line =~ m/<text>(.*)<\/text>/ )
    {
        # Replace the old meta text with the new one.
        $metaText = $1;
    }
    else
    {
        # Normal line. The old meta text applies.
        
        my @words = split /\s+/ $line;
        
        foreach my $word (@words)
        {
            store_in_index($metaText, $word);
        }
    }
}

close $srcFile;
[download]

In the example above, the store_in_index() function could anything from storage into a simple hash, to a relational database. It would depend on how much data, and how long you want to keep that data.

An alternative approach would be one where you read the file backwards. Firstly for the word the user is looking for, and then for the meta information line that it relates to. Off the top of my head, I am not sure how that would be done, but I am sure there are ways.

File::Bidirectional / File::ReadBackwards

Instead of first searching and then looking for the tag you could read the file so that you already have separated each text and its tag

For example you can change $/ to "<text>" so that when you read that file, you read it not line by line, but text by text.

Or just parse line by line and store into a new index of an array whenever a line begins with '<text>'. Have an array for the tags and an array for the texts where corresponding tag and text have the same index and you have all the information "finely sliced at your fingertips" ;-)

my @something;
{
    my $before = [];
    my $after  = [];
    my $curr   = $before;
    my $meta   = [];

  LOOP:
    while (<$fh>) {
        if (/^\Q..../) {
            $curr = $after;
            next LOOP;
        }

        if (/^</) {
## Something( { } );
            push @something, {
                meta   => $meta,
                before => $before,
                after  => $after,
              };

            $before = [];
            $after  = [];
            $curr   = $before;
            $meta   = [$_];

            next LOOP;
        } ## end if (/^</)

        push @$curr, $_;
    } ## end while (<$fh>)

    if ( grep { $_ or @$_ } $meta, $before, $after ) {
        push @something, {
            meta   => $meta,
            before => $before,
            after  => $after,
          };
    } ## end if ( grep { $_ or @$_ ...})

}
[download]

shows in a TK::Text Widget the portion of the text containing that word. This works fine.

So hook into the first part: as it must have searched for the start, all you have to do is retrieve the contents:

my($meta) = substr($everything, $sectionstart) =~ /<text>(.*?)<\/text>
+/;
[download]

echo secret > fred.txt:hidden
more < fred.txt:hidden
secret
[download]

while (<FILE>){
if (/<text>(.*)<\/text>/) {$currentheader = $1} # if the current line 
+has a header, save it, otherwise, keep last saved header
# your data search code goes here
if (there was a hit) {print $currentheader along with the hit}
}
[download]

Peace of code, is that like peace of mind? If peace doesn't inspire you, maybe you're not looking for inspiration?