tobyink has provided an excellent solution. For your future reference--and in case the need arises again--there are Perl modules that can be used for parsing the kind of text you have. Here's an example that uses Mojo::DOM to parse your <a> tags:
use strict;
use warnings;
use Mojo::DOM;
my $text = <<END;
<a>
word1
word2
word3
</a>
<a>
word4
word5
</a>
<a>
word6
word7
</a>
END
my $dom = Mojo::DOM->new($text);
my $i = 1;
for my $chunk ( $dom->find('a')->each ) {
print 'Chunk ' . $i++ . ': ' . $chunk->text . "\n";
}
Output:
Chunk 1: word1 word2 word3
Chunk 2: word4 word5
Chunk 3: word6 word7
Thus, each group that you need to analyze is contained by $chunk->text within the for loop.
Hope this helps!