This could be entered into a hash fairly simply. First set the filehandle to read in paragraph mode, (local $/ = ''), within a limited scope (determined by the enclosing braces, {  }). Then a simple split /\n/ will capture the key and the values in an array.

EDIT: added for loop that removes quotes from beginning and end of items in @data

Usage: perl my_program.pl input_file

#!/usr/bin/perl
use strict;
use warnings;
use Data::Dumper;

my %data;

{
    local $/ = '';
    while (<>) {
        my ($key, @data) = split /\n/;
        for (@data) {
             s/^"//;
             s/"$//;
        }
        $data{$key} = \@data;
    }   
}

$Data::Dumper::Indent=1;

print Dumper \%data;
[download]

Prints:

$VAR1 = {
  'arn:aws:iam::12345678901:role/Role2-Role2' => [
    'Alexa for Business',
    'AWS Certificate Manager',
    'AWS Certificate Manager Private Certificate Authority',
    'AWS Amplify',
    'Manage - Amazon API Gateway',
    'Application Auto Scaling',
    'AWS App Mesh',
    'Amazon AppStream 2.0',
    'AWS AppSync'
  ],
  'arn:aws:iam::11111111111:role/ADFS-MyRoleName' => [
    'Alexa for Business',
    'AWS Certificate Manager',
    'AWS Certificate Manager Private Certificate Authority',
    'AWS Amplify',
    'Manage - Amazon API Gateway',
    'AWS App Mesh',
    'Amazon AppStream 2.0',
    'AWS AppSync',
    'Amazon Athena',
    'AWS Auto Scaling'
  ]
};
[download]

(The output is a little messy due to the long key names.)

EDIT: Used Indent to Data::Dumper to clean up output from Dumper as noted by jdporter