Lingua::ZH::Toke - Chinese Tokenizer


This document describes version 0.02 of Lingua::ZH::Toke, released January 11, 2004.


    use Lingua::ZH::Toke;

    # -- if inputs are unicode strings, use the two lines below instead
    # use utf8;
    # use Lingua::ZH::Toke 'utf8';

    # Create Lingua::ZH::Toke::Sentence object (->Sentence also works)
    my $token = Lingua::ZH::Toke->new( '那人卻在/燈火闌珊處/益發意興闌珊' );

    # Easy tokenization via array deferencing
    print $token->[0]           # Fragment       - 那人卻在
                ->[2]           # Phrase         - 卻在
                ->[0]           # Character      - 卻
                ->[0]           # Pronounciation - ㄑㄩㄝˋ
                ->[2];          # Phonetic        - ㄝ

    # Magic histogram via hash deferencing
    print $token->{'那人卻在'};     # 1 - One such fragment there
    print $token->{'意興闌珊'};     # 1 - One such phrase there
    print $token->{'發意興闌'};     # undef - That's not a phrase
    print $token->{'珊'};        # 2 - Two such character there
    print $token->{'ㄧˋ'};       # 2 - Two such pronounciation: 益意
    print $token->{'ㄨ'};        # 3 - Three such phonetics: 那火處

    # Iteration over fragments
    while (my $fragment = <$token>) {
        # Iteration over phrases
        while (my $phrase = <$fragment>) {
            # ...


This module puts a thin wrapper around Lingua::ZH::TaBE, by blessing refereces to TaBE's objects into its English counterparts.

Besides offering more readable class names, this module also offers various overloaded methods for tokenization; please see "SYNOPSIS" for the three major ones.

Since Lingua::ZH::TaBE is a Big5-oriented module, we also provide a simple utf8 layer around it; if you have Perl version 5.6.1 or later, just use this:

    use utf8;
    use Lingua::ZH::Toke 'utf8';

With the utf8 flag set, all Toke objects will stringify to unicode strings, and constructors will take either unicode strings, or big5-encoded bytestrings.

Note that on Perl 5.6.x, Encode::compat is needed for the utf8 feature to work.


The constructor methods correspond to the six object levels: ->Sentence, ->Fragment, ->Phrase, ->Character, ->Pronounciation and ->Phonetic. Each of them takes one string argument, representing the string to be tokenized.

The ->new method is an alias to ->Sentence>.

All object methods, except ->new, are passed to the underlying Lingua::ZH::TaBE object.


Under utf8 mode, you may sometimes need to explicitly stringify the return values, so their utf8 flag can be properly set:

    $value = $token->[0];       # this may or may not work
    $value = "$token->[0]";     # this is guaranteed to work

This module does not care about efficiency or memory consumption yet, hence it's likely to fail miserably if you demand either of them. Patches welcome.

As the name suggests, the chosen interface is very bizzare. Use it at the risk of your own sanity.


Lingua::ZH::TaBE, Encode::compat, Encode


Autrijus Tang <>


Copyright 2003, 2004 by Autrijus Tang <>.

This program is free software; you can redistribute it and/or modify it under the same terms as Perl itself.