I agree with Fletch above. This uses HTML::TokeParser::Simple. It pushes all the anchor hrefs and anchor text onto an array. You can then easily choose which you need.

imo this is much easier and more reliable than trying to build, debug and maintain a complex regex on something as loose as HTML.

It's also easy to adapt if the spec changes (and doesn't it always) or to other parsing tasks as they arise.

I believe using a parser (and there are many to suit all tastes) is a big win on all counts and I highly recommend it.

#!/usr/bin/perl

use strict;
use warnings;
use Data::Dumper;
use HTML::TokeParser::Simple;

my $html;
{
  local $/;
  $html = <DATA>
}

my $p = HTML::TokeParser::Simple->new(\$html);
$p->unbroken_text(1);

my ($in_li, @record, @db);

while (my $t = $p->get_token){
  
  $in_li++, next if $t->is_start_tag('li');
  next unless $in_li;
  
  if ($t->is_end_tag('li')){
    push @db, [@record];
    $in_li = 0;
    next;
  }
  
  if ($t->is_start_tag('a')){
    push @record, $t->get_attr('href');
    my $text = $p->get_trimmed_text('/a');
    push @record, $text;
  }
  
}

#die Dumper \@db;

# the text inside the first link's text, the 2nd link's URL, the 2nd l
+ink's text.

for my $record (@db){
  my @field = @{$record};
  print $field[1], "::", $field[2], "::", $field[3], "\n";
}


__DATA__
<li>
  <a class="style5" href="http://www.site.com/page.html">
    some words here
  </a>
    - 
  <a class="style3" href="http://www.site.com/page2.html">
    "some words here"
  </a> 
</li>
[download]

output:

---------- Capture Output ----------
> "C:\Perl\bin\perl.exe" _new.pl

some words here::http://www.site.com/page2.html::"some words here"

> Terminated with exit code 0.
[download]