Sentence::tokenize() now uses flex-based ::tokenize()

[vspell.git] / scripts / personal-name-extract-from-hcmuns.pl

#!/usr/bin/perl

%entity = (
   'à'  => 'à',
   'ả' => 'ä',
   'ã'  => 'ã',
   'á'  => 'á',
   'ạ' => 'Õ',
   'ă'  => 'å',
   'ằ' => '¢',
   'ẳ' => 'Æ',
   'ẵ' => 'Ç',
   'ắ' => '¡',
   'ặ' => '£',
   'â'  => 'â',
   'ầ' => '¥',
   'ẩ' => '¦',
   'ẫ' => 'ç',
   'ấ' => '¤',
   'ậ' => '§',
   'è'  => 'è',
   'ẻ' => 'ë',
   'ẽ' => '¨',
   'é'  => 'é',
   'ẹ' => '©',
   'ê'  => 'ê',
   'ề' => '«',
   'ể' => '¬',
   'ễ' => '­',
   'ế' => 'ª',
   'ệ' => '®',
   'ì'  => 'ì',
   'ỉ' => 'ï',
   'ĩ'  => 'î',
   'í'  => 'í',
   'ị' => '¸',
   'ò'  => 'ò',
   'ỏ' => 'ö',
   'õ'  => 'õ',
   'ó'  => 'ó',
   'ọ' => '÷',
   'ô'  => 'ô',
   'ồ' => '°',
   'ổ' => '±',
   'ỗ' => '²',
   'ố' => '¯',
   'ộ' => 'µ',
   'ơ'  => '½',
   'ờ' => '¶',
   'ở' => '·',
   'ỡ' => 'Þ',
   'ớ' => '¾',
   'ợ' => 'þ',
   'ù'  => 'ù',
   'ủ' => 'ü',
   'ũ'  => 'û',
   'ú'  => 'ú',
   'ụ' => 'ø',
   'ư'  => 'ß',
   'ừ' => '×',
   'ử' => 'Ø',
   'ữ' => 'æ',
   'ứ' => 'Ñ',
   'ự' => 'ñ',
   'ỳ' => 'Ï',
   'ỷ' => 'Ö',
   'ỹ' => 'Û',
   'ý'  => 'ý',
   'ỵ' => 'Ü',
   'đ'  => 'ð',
   'À'  => 'À',
   'Ả' => 'Ä',
   'Ã'  => 'Ã',
   'Á'  => 'Á',
   'Ạ' => '€',
   'Ă'  => 'Å',
   'Ằ' => '‚',
   'Ẳ' => '\x02',
   'Ẵ' => '\x05',
   'Ắ' => '�',
   'Ặ' => 'ƒ',
   'Â'  => 'Â',
   'Ầ' => '…',
   'Ẩ' => '†',
   'Ẫ' => '\x06',
   'Ấ' => '$',
   'Ậ' => '‡',
   'È'  => 'È',
   'Ẻ' => 'Ë',
   'Ẽ' => 'ˆ',
   'É'  => 'É',
   'Ẹ' => '‰',
   'Ê'  => 'Ê',
   'Ề' => '‹',
   'Ể' => 'Œ',
   'Ễ' => '�',
   'Ế' => 'Š',
   'Ệ' => 'Ž',
   'Í'  => 'Í',
   'Ỉ' => '›',
   'Ĩ'  => 'Î',
   'Ì'  => 'Ì',
   'Ị' => '˜',
   'Ò'  => 'Ò',
   'Ỏ' => '™',
   'Õ'  => ' ',
   'Ó'  => 'Ó',
   'Ọ' => 'š',
   'Ô'  => 'Ô',
   'Ồ' => '�',
   'Ổ' => '‘',
   'Ỗ' => '’',
   'Ố' => '�',
   'Ộ' => '“',
   'Ơ'  => '´',
   'Ờ' => '–',
   'Ở' => '—',
   'Ỡ' => '³',
   'Ớ' => '•',
   'Ợ' => '”',
   'Ù'  => 'Ù',
   'Ủ' => 'œ',
   'Ũ'  => '�',
   'Ú'  => 'Ú',
   'Ụ' => 'ž',
   'Ư'  => '¿',
   'Ừ' => '»',
   'Ử' => '¼',
   'Ữ' => 'ÿ',
   'Ứ' => 'º',
   'Ự' => 'q',
   'Ỳ' => 'Ÿ',
   'Ỷ' => '\x14',
   'Ỹ' => '\x19',
   'Ý'  => 'Ý',
   'Ỵ' => '\x1e',
   'Đ'  => 'Ð',
);

# Get all content
while (<>) { 
        chomp; 
        if (m,                <td align="center">[0-9]*</td>,) {
                $url = <>;
                $surname = <>;chomp($surname);
                $name = <>;chomp($name);
                $surname =~ s,.*<td>(.*)</td>.*,\1,;
                $name =~ s,.*<td>(.*)</td>.*,\1,;
                $name =~ s,\s*$,,;
                $surname =~ s,\s*$,,;
                $line =  "$name, $surname\n";
                foreach $i (keys %entity) {
                    $val = $entity{$i};
                    $line =~ s/$i/$val/g;
                }
                print $line;
        }
}

# Local Variables:
# coding: viscii
# End: