use strict;
use warnings;
use Data::Dumper;

my %mycorpus = (
          text1 => "<p><time datetime=2017-09-04T05:23:39Z>04/09/17 06
+:23:39</time>
                    Irrelevant text that I do not need.
##a##
gt##a b c##
##a b a c d A b##
 <97> 164 notes",

            text2 => "p><time datetime=2017-09-30T18:20:56Z>30/09/17 1
+9:20:56</time>
            Irrelevant text that may feature the word soft, softest, o
+r softly.
##a##
##a b a ##

 <97> 379 notes
            Irrelevant text.",

            text3 => "<p><time datetime=2017-09-30T19:27:03Z>30/09/17 
+19:27:03</time>
##C##
##A##
##b##
 <97> 180 notes
 Irrelevant text."
);

my %counts;
my %overallcounts;

foreach my $filename (sort keys %mycorpus) {
        my $date;
        my $hashtags = '';

        #find the dates
        if ($mycorpus{$filename} =~ /(\d{4}-\d{2}-\d{2})T/g){
            $date = $1;
        }
        
        #find only the relevant text
        while ($mycorpus{$filename} =~ /##(.*)##/g){
            $hashtags = $1;
            
        #split text into words
            my @words = split /\W+/, $hashtags;
            
            foreach my $word (@words){
                
                if ($word =~ /(\w+)/gi){
                    $word =~ tr/A-Z/a-z/;
                    $counts{$date}{$word}++;
                    $overallcounts{$date}++; #new hash to help with re
+lative frequency overall count per day
                }
            }           
        }        
}

print Dumper \%counts;
[download]

The current output is as follows:

$VAR1 = {
          '2017-09-04' => {
                            'b' => 3,
                            'd' => 1,
                            'c' => 2,
                            'a' => 5
                          },
          '2017-09-30' => {
                            'b' => 2,
                            'c' => 1,
                            'a' => 4
                          }
        };
[download]

However, my expected output would be something like this (I've rounded to 2 d.p, but this is not necessary for the final results).

$VAR1 = {
          '2017-09-04' => {
                            'b' => 27.27,
                            'd' => 9.09,
                            'c' => 18.18,
                            'a' => 45.45
                          },
          '2017-09-30' => {
                            'b' => 28.57,
                            'c' => 14.29,
                            'a' => 57.14
                          }
        };
[download]

However, I am really struggling to work out how to do this. I can't picture how this script would look and I've been unable to find any relevant examples online.

I think that I might need to utilise a second hash, where the dates are the keys and the total number of words per day are the values (and I've done this in the example above). I would then need to divide each of the values in my original hash of hashes by the corresponding values in this new hash, and then multiply this new number by 100. The returned value would then need to be stored in my hash of hashes.

However, while I know the steps that I need to take in theory, I have really reached my current 'Perl-knowledge' limit, and I would really appreciate any advice on how to implement this final part of my code.