I demo a common parsing pattern below. You figure out what is special about the start of a "new record". If you see that "special thing" and you are already working on a record, then you process the previous record and start a new one. Otherwise you are continuing the current record. Note that since the start of a new record triggers the output of the previous record, there is a need to output the final record once the data ends.

use strict;
use warnings;
$|=1;

my $data_lines = 
('keyword1 data1 data2 data3
keyword2 data1 data2 data3
   data4 data5
   data6
keyword1 data1 data2 data3 data4
keyword3 data1
');

my @lines = split (/\n/,$data_lines);
print "To show array of text lines as per spec:\n";

foreach (@lines)
{
   print "  $_\n";
}
print "\n";

print "Showing data array's per combined input lines:\n\n";

my @array = ();
foreach my $line (@lines)
{
    if ($line =~ /^\S/ and @array>0) # Finish previous record
    {  
         process_array (@array);
         @array = ();   #start new record
         push (@array,$_) foreach (split ' ',$line);
    }
    else   # new or continuing record
    {
        push (@array, $_) foreach (split ' ',$line);   
    }    
}
process_array (@array); # the last record

sub process_array
{
   my @array = @_;
   print "process array in some sub = @array\n";
}   

__END__
To show array of text lines as per spec:
  keyword1 data1 data2 data3
  keyword2 data1 data2 data3
     data4 data5
     data6
  keyword1 data1 data2 data3 data4
  keyword3 data1

Showing data array's per combined input lines:

process array in some sub = keyword1 data1 data2 data3
process array in some sub = keyword2 data1 data2 data3 data4 data5 dat
+a6
process array in some sub = keyword1 data1 data2 data3 data4
process array in some sub = keyword3 data1
[download]