diff options
author | Chris Dyer <cdyer@allegro.clab.cs.cmu.edu> | 2013-01-15 01:20:00 -0500 |
---|---|---|
committer | Chris Dyer <cdyer@allegro.clab.cs.cmu.edu> | 2013-01-15 01:20:00 -0500 |
commit | 9d7167751a3712a79ad356764d803106a71ce5e3 (patch) | |
tree | 4eb2feff7b0c6bb66ca28f70f45a933fc2aed600 /corpus | |
parent | a52035d1bce0566bfd9603d93e4fb601bf25a73f (diff) |
corpus files
Diffstat (limited to 'corpus')
-rwxr-xr-x | corpus/add-self-translations.pl | 2 | ||||
-rwxr-xr-x | corpus/filter-length.pl | 6 | ||||
-rwxr-xr-x | corpus/paste-files.pl | 12 |
3 files changed, 16 insertions, 4 deletions
diff --git a/corpus/add-self-translations.pl b/corpus/add-self-translations.pl index 153bc454..d707ce29 100755 --- a/corpus/add-self-translations.pl +++ b/corpus/add-self-translations.pl @@ -6,7 +6,7 @@ use strict; my %df; my %def; while(<>) { - print; +# print; chomp; my ($sf, $se) = split / \|\|\| /; die "Format error: $_\n" unless defined $sf && defined $se; diff --git a/corpus/filter-length.pl b/corpus/filter-length.pl index 70032ca7..3cfa40cc 100755 --- a/corpus/filter-length.pl +++ b/corpus/filter-length.pl @@ -3,8 +3,8 @@ use strict; use utf8; ##### EDIT THESE SETTINGS #################################################### -my $MAX_LENGTH = 99; # discard a sentence if it is longer than this -my $AUTOMATIC_INCLUDE_IF_SHORTER_THAN = 6; # if both are shorter, include +my $MAX_LENGTH = 150; # discard a sentence if it is longer than this +my $AUTOMATIC_INCLUDE_IF_SHORTER_THAN = 7; # if both are shorter, include my $MAX_ZSCORE = 1.8; # how far from the mean can the (log)ratio be? ############################################################################## @@ -128,6 +128,8 @@ while(<F>) { next; } print; + } else { + print; } $to++; } diff --git a/corpus/paste-files.pl b/corpus/paste-files.pl index 24c70599..0b788386 100755 --- a/corpus/paste-files.pl +++ b/corpus/paste-files.pl @@ -17,6 +17,7 @@ for my $file (@ARGV) { binmode(STDOUT,":utf8"); binmode(STDERR,":utf8"); +my $bad = 0; my $lc = 0; my $done = 0; my $fl = 0; @@ -34,7 +35,15 @@ while(1) { last; } chomp $r; - die "$ARGV[$anum]:$lc contains a ||| symbol - please remove.\n" if $r =~ /\|\|\|/; + if ($r =~ /\|\|\|/) { + $r = ''; + $bad++; + } + warn "$ARGV[$anum]:$lc contains a ||| symbol - please remove.\n" if $r =~ /\|\|\|/; + $r =~ s/\|\|\|/ /g; + $r =~ s/ +//g; + $r =~ s/^ //; + $r =~ s/ $//; $anum++; push @line, $r; } @@ -47,4 +56,5 @@ for (my $i = 1; $i < scalar @fhs; $i++) { my $r = <$fh>; die "Mismatched number of lines.\n" if defined $r; } +print STDERR "Bad lines containing ||| were $bad\n"; |