11 Mar 2006 05:47:55 UTC
- Distribution: KinoSearch
- Source (raw)
- Browse (raw)
- How to Contribute
- Issues (5)
- Testers (4 / 3 / 0)
- KwaliteeBus factor: 0
- License: perl_5
- Activity24 month
- Download (169.28KB)
- MetaCPAN Explorer
- Subscribe to distribution
- This version
- Latest versionCREAMYG Marvin Humphreyand 1 contributors
- Marvin Humphrey <marvin at rectangular dot com>
KinoSearch::Analysis::Tokenizer - customizable tokenizing
my $whitespace_tokenizer = KinoSearch::Analysis::Tokenizer->new( token_re => qr/\S+/, ); # or... my $word_char_tokenizer = KinoSearch::Analysis::Tokenizer->new( token_re => qr/\w+/, ); # or... my $apostrophising_tokenizer = KinoSearch::Analysis::Tokenizer->new; # then... once you have a tokenizer, put it into a PolyAnalyzer my $polyanalyzer = KinoSearch::Analysis::PolyAnalyzer->new( analyzers => [ $lc_normalizer, $word_char_tokenizer, $stemmer ], );
Generically, "tokenizing" is a process of breaking up a string into an array of "tokens".
# before: my $string = "three blind mice"; # after: @tokens = qw( three blind mice );
KinoSearch::Analysis::Tokenizer decides where it should break up the text based on the value of
# before: my $string = "Eats, Shoots and Leaves."; # tokenized by $whitespace_tokenizer @tokens = qw( Eats, Shoots and Leaves. ); # tokenized by $word_char_tokenizer @tokens = qw( Eats Shoots and Leaves );
my $tokenizer = KinoSearch::Analysis::Tokenizer->new( token_re => $matches_one_token, );
Construct a Tokenizer object.
token_re must be a pre-compiled regular expression matching one token. It must not use any capturing parentheses, though non-capturing parentheses are fine:
# match "O'Henry" as well as "Henry" and "it's" as well as "it" my $token_re = qr/ \b # start with a word boundary \w+ # Match word chars. (?: # Group, but don't capture... '\w+ # ... an apostrophe plus word chars. )? # Matching the apostrophe group is optional. \b # end with a word boundary /xsm; my $apostrophizing_tokenizer = KinoSearch::Analysis::Tokenizer->new( token_re => $token_re, );
Incidentally, the above token_re is the default value.
Copyright 2005-2006 Marvin Humphrey
See KinoSearch version 0.08.