NAME
Lingua::JA::TFWebIDF - TF*WebIDF calculator
SYNOPSIS
use Lingua::JA::TFWebIDF;
use utf8;
use feature qw/say/;
use Data::Printer;
my $tfidf = Lingua::JA::TFWebIDF->new(
pos1_filter => [qw/非自立 代名詞 ナイ形容詞語幹 副詞可能 サ変接続/],
ng_word => [qw/編集 本人 自身 自分 たち さん 年 月 日/],
);
my %tf = (
'自然言語処理' => 9,
'自然言語' => 6,
'自然言語理解' => 4,
'処理' => 5,
'解析' => 4,
);
p $tfidf->tfidf(\%tf)->dump;
p $tfidf->tfidf($text)->dump;
p $tfidf->tf($text)->dump;
for my $result (@{ $tfidf->tfidf($text)->list(20) })
{
my ($word, $weight) = each %{$result};
say "$word: $weight";
}
DESCRIPTION
Lingua::JA::TFWebIDF calculates TF*WebIDF weight.
Compared with Lingua::JA::TFIDF, this module has the following advantages.
supports Tokyo Cabinet and many options.
tfidf method accepts \%tf. (This facilitates the use of other morphological analyzers.)
METHODS
new( %config || \%config )
Creates a new Lingua::JA::TFWebIDF instance.
The following configuration is used if you don't set %config.
KEY DEFAULT VALUE
----------- ---------------
pos1_filter [qw/非自立 代名詞 ナイ形容詞語幹 副詞可能/]
pos2_filter []
pos3_filter []
ng_word []
term_length_min 2
term_length_max 30
concat_max 30
tf_min 1
df_min 0
df_max 250_0000_0000
fetch_unk_word_df 0
db_auto 1
guess_df 1
idf_type 1
api 'YahooPremium'
appid undef
driver 'TokyoCabinet'
df_file './df.tch'
fetch_df 0
expires_in 365
documents 250_0000_0000
Furl_HTTP undef
verbose 0
- pos(1|2|3)_filter => \@pos
-
The filters of '品詞細分類'.
- concat_max => $num
-
The maximum value of the number of term concatenations.
If 2 is specified, 2 consecutive nouns are concatenated. I recommend that you specify a large value or 0.
- fetch_df => 0 || 1
-
If df_file has the (Web)DF and it has not expired yet, it is used.
If it is not so,
1: fetches the DF of the word which exists in the dictionary of MeCab.
0: if guess_df is 1, guesses the DF.
if guess_df is 0, does not give TF*WebIDF weight.
- fetch_unk_word_df => 0 || 1
-
'unk word' is a word which not exists in the dictionary of MeCab.
If df_file has the (Web)DF and it has not expired yet, it is used.
If it is not so,
1: fetches the DF of the unk word if fetch_df is 1.
0: if guess_df is 1, guesses the DF.
if guess_df is 0, does not give TF*WebIDF weight.
- db_auto => 0 || 1
-
If 1 is specified, (open|close)s the WebDF(Document Frequency) database automatically. This option works when tfidf method is called.
- guess_df => 0 || 1
-
If df_file has the (Web)DF and it has not expired yet, it is used.
If it is not so,
1: guesses the DF based on the string length of $word.
0: doesn't give TF*WebIDF weight.
- idf_type, api, appid, driver, df_file, expires_in, documents, Furl_HTTP, verbose
-
See Lingua::JA::WebIDF.
tfidf( $text || \%tf )
Calculates TF*WebIDF weight. If scalar value is set, MeCab separates the value into appropriate morphemes. If you want to use other morphological analyzers, you have to set a hash reference which contains terms and their TF(Term Frequency).
tf($text)
Calculates TF(Term Frequency) via MeCab.
idf, df, db_open, db_close, purge
See Lingua::JA::WebIDF.
AUTHOR
pawa <pawapawa@cpan.org>
SEE ALSO
Lingua::JA::WebIDF::Driver::TokyoTyrant
LICENSE
This library is free software; you can redistribute it and/or modify it under the same terms as Perl itself.