NAME

Lingua::JA::WebIDF - WebIDF calculator

SYNOPSIS

  use Lingua::JA::WebIDF;

  my $webidf = Lingua::JA::WebIDF->new
  (
      api       => 'Yahoo',
      appid     => $appid,
      fetch_df  => 1,
      Furl_HTTP => { timeout => 3 }
  );

  print $webidf->idf("東京"); # low
  print $webidf->idf("スリジャヤワルダナプラコッテ"); # high

DESCRIPTION

Lingua::JA::WebIDF calculates WebIDF scores.

WebIDF(Inverse Document Frequency) scores represent the rarity of words on the Web. The WebIDF scores of rare words are high. Conversely, the WebIDF scores of common words are low.

IDF is based on the intuition that a query term which occurs in many documents is not a good discriminator and should be given less weight than one which occurs in few documents.

METHOD

new( %config || \%config )

Creates a new Lingua::JA::WebIDF instance.

The following configuration is used if you don't set %config.

  KEY                 DEFAULT VALUE
  -----------         ---------------
  idf_type            1
  api                 'Yahoo'
  appid               undef
  driver              'Storable'
  df_file             undef
  fetch_df            1
  expires_in          365
  documents           250_0000_0000
  default_df          5000
  Furl_HTTP           undef

idf_type => 1 || 2 || 3

The type1 is the most commonly cited form of IDF.

                   N
  idf(t_i) = log -----  (1)
                  n_i

  N  : the number of documents
  n_i: the number of documents which contain term t_i
  t_i: term

The type2 is a simple version of the RSJ weight.

              N - n_i + 0.5
  w_i = log ----------------  (2)
               n_i + 0.5

The type3 is a modification of (2).

              N + 0.5
  w_i = log -----------  (3)
             n_i + 0.5

api => 'Yahoo' || 'YahooPremium' || 'Bing'

Uses the specified Web API when fetches WebDF(Document Frequency) scores from the Web.

driver => 'Storable' || 'TokyoCabinet'

Fetches and saves WebDF scores with the specified driver.

df_file => $path

Saves WebDF scores to the specified path.

If undef is specified, 'yahoo_utf8.st' is used. This file is located in 'Lingua/JA/WebIDF/' and contains the WebDF scores of about 60000 words. There are other format files in the 'df' directory of this library.

I recommend that you change the file depending on the type of Web API you specifies because WebDF may be different depending on it.

fech_df => 0

Doesn't fetch WebDF scores. (If 0 is specified.)

If the WebDF score you want to know is already saved, it is used. Otherwise, the value of default_df is used.

expires_in => $days

If 365 is specified, a WebDF score expires in 365 days after fetches it.

Furl_HTTP => \%option

Sets the options of Furl::HTTP->new.

If you want to use proxy server, you have to use this option.

idf($word)

Calculates the WebIDF score of $word.

If the WebDF score of $word is not saved or is expired, fetches it by using the Web API you specified and saves it.

df($word)

Fetches the WebDF score of $word.

If the WebDF score of $word is not saved or is expired, fetches it by using the Web API you specified and saves it.

AUTHOR

pawa <pawapawa@cpan.org>

SEE ALSO

Lingua::JA::TFIDF

Bing API: http://www.bing.com/toolbox/bingdeveloper/

Yahoo API: http://developer.yahoo.co.jp/

Tokyo Cabinet: http://fallabs.com/tokyocabinet/

S. Robertson, Understanding inverse document frequency: on theoretical arguments for IDF. Journal of Documentation 60, 503-520, 2004.

LICENSE

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