Pratheepan Raveendranathan

NAME

WebService::GoogleHack - Perl package that ties together all GoogleHack modules (Webservice::GoogleHack::Search, Webservice::GoogleHack::Spelling, Webservice::GoogleHack::Rate, and Webservice::GoogleHack::Text) to implement Natural Language Processing techniques that use the World Wide Web as a source of information. Use this package to access all the functionality of GoogleHack.

SYNOPSIS

    use WebService::GoogleHack;

    my $google = new WebService::GoogleHack;

    #Initializing the object to the contents of the configuration file
    # API Key, GoogleSearch.wsdl file location.

    $google->initConfig("initconfig.txt");

    #Printing the contents of the configuration file
    $google->printConfig();

    #Measure the semantic relatedness between the words "white house" and 
    #"president".

    $measure=$google->measureSemanticRelatedness1("white house","president");

    print "\nRelatedness measure between white house and president is: ";
    print $measure."\n";

    #Going to search for words that are related to "toyota" and "ford" 
    my @terms=();
    push(@terms,"toyota");
    push(@terms,"ford");

    #The parameters are the search terms, number of web page results to look 
    #at, the number of iterations,output file and the "true" indicates that the
    #diagnostic data should be stored in the file "results.txt"

    $results=$google->Algorithm1(\@terms,10,25,1,"results.txt","true");

    print $results;

DESCRIPTION

WebService::GoogleHack is a PERL package that interacts with the Google API, and implements basic functions that allow the user to interact with Google and retrieve results in an easy to use format. GoogleHack also implements and extends a number of Natural Language Processing by using the World Wide Web as a source of information.

Some of the features are:

    * Issue queries to Google (WebService::GoogleHack, WebService::GoogleHack::Search)

    * Retrieve Spelling suggestions from Google (WebService::GoogleHack, WebService::GoogleHack::Spelling)

    * Find the Pointwise Mututal Information (PMI) measure between two words (WebService::GoogleHack,WebService::GoogleHack::Rate)

    * Given a paragraph find if the paragraph has a positive or negative semantic orientation.(WebService::GoogleHack,WebService::GoogleHack::Rate)
         
    * Given a set of words along with a positively oriented word such as "excellent" and a negatively oriented 
      word such as "poor", find if the word has a positive or negative semantic orientation.(WebService::GoogleHack,WebService::GoogleHack::Rate)

    * Given a set of phrases along with a positively oriented word such  as "excellent" and a negatively oriented word 
      such as "poor", predict if the given phrases are positive or negative in sentiment.(WebService::GoogleHack,WebService::GoogleHack::Rate)

    * Given two or more words finds a set of related words. (WebService::GoogleHack)


 

WebService::GoogleHack::Text - GoogleHack uses this module to manipulate text retrieved from the web (Get n-word sentences, words,and parse HTML etc).

WebService::GoogleHack::Search - GoogleHack uses this module to query Google.

WebService::GoogleHack::Rate - GoogleHack users this module to implement some of the Sentiment Classification algorithms.

WebService::GoogleHack::Spelling - GoogleHack uses this module to query Google for spelling suggestions.

REQUIRED PACKAGES

1) Google API (http://www.google.com/apis/)

2) Brill Tagger (If using Sentiment Classification stuff)

    Installation file and instructions @ : 
   
    http://www.cs.jhu.edu/~brill/RBT1_14.tar.Z

    Instructions also available in GoogleHack INSTALL file.

3) Required PERL Modules

    SOAP::Lite;

    Set::Scalar;

    Text::English;

    LWP::Simple;

    URI::URL;

    LWP::UserAgent;

    HTML::LinkExtor;
 
    Data::Dumper;

FUNCTIONS

GENERAL FUNCTIONALITY

These are the GoogleHack functions that are common to all sort of operations. These functions are used to create and initialize GoogleHack objects.

__METHOD__->new()

Purpose: This function creates an object of type GoogleHack and returns a blessed reference.

returns: A blessed reference to a GoogleHack object.

__METHOD__->initConfig(configLocation)

Purpose: This function is used to read the configuration file containing information such as the Google-API key, the base directory path, and the path to the Brill Tagger. The configuration file is in the WebService/GoogleHack/Datafiles directory.

This function must be called in order to initialize the GoogleHack object.

Valid arguments are :

  • configLocation

    string. Location of the configuration file.

returns : Returns an object which contains the parsed information.

__METHOD__->printConfig()

Purpose: This function is used to print the information read from the configuration file

No arguments.

This set of functions deal with the problem of finding sets of related words by using the World Wide Web as a source of information.

__METHOD__->measureSemanticRelatedness1(searchString1,searchString2)

Purpose: This function is used to measure the relatedness between two words.

Formula used: log(hits(w1)) + log(hits(w2)) - 2 * log(hits(w1w2))

Valid arguments are :

  • searchString1

    string. The search string which can be a phrase or word

  • searchString2

    string. The search string which can be a phrase or word

Returns: Returns the object containing the relatedness measure.

__METHOD__->measureSemanticRelatedness2(searchString1,searchString2)

Purpose: This function is used to measure the relatedness between two words.

Formula used: log(w1w2/(w1+w2))

Valid arguments are :

  • searchString1

    string. The search string which can be a phrase or word

  • searchString2

    string. The search string which can be a phrase or word

Returns: Returns the object containing the relatedness measure.

__METHOD__->measureSemanticRelatedness3(searchString1,searchString2)

Purpose: This function is used to measure the relatedness between two words.

Formula used: log( hits(w1w2) / (hits(w1) * hits(w2)))

Valid arguments are :

  • searchString1

    string. The search string which can be a phrase or word

  • searchString2

    string. The search string which can be a phrase or word

Returns: Returns the object containing the relatedness measure.

__METHOD__->Algorithm1(searchTerms,N,C,I,trace, html)

Purpose:Given two or more words, this function tries to find a set of related words. This is the Google-Hack baseline algorithm 1. For example, given the two words gun and pistol, an example of an expanded set of related words would be,

{laser,paintball, case,bullet, machine gun, rifle} etc.

  The features of Initial Approach (Algorithm 1) is given below

                  - Frequency Based
 
                  - Accepts only 2 terms

                  - Results also contain only unigrams
                
                  - A frequency cutoff is used
                
                  - Stop words and web stop words are removed.
  • searchTerms

    string. The array of search terms (Can only be a word). =item *

    N

    number. The number of web pages results to be looked at.

  • C

    number. Words occuring less than the frequencyCutoff would be excluded from results.

  • I

    number. The number of iterations that you want the function to search and build cluster on.

  • trace.

    string. The location where the file containing the retreived information has to be stored.

  • html.

    bool. If set to "true" then the results returned by the algorithm is in HTML format. If not "true", the results are in plain text format.

returns : Returns an html or text version of the results.

__METHOD__->Algorithm2(searchTerms,N,C,BC, I,S,SC,trace, html)

Purpose:Given two or more words, this function tries to find a set of related words. This is the Google-Hack algorithm 2.

   The features of Second Approach (Algorithm 2) is given below

                  - Accepts more than 2 terms

                  - Has 3 relatedness scores
                
                  - Accepts unigrams and 2-word collocation as input
                
                  - Results also contain 2-word collocations
                
                  - A score cutoff is also included along with frequency cutoff
                
                  - A bigram cutoff is also included.

                  - Stop words and web stop words are removed.

                  - Stop phrases and web stop phrases are removed.
  • searchTerms

    string. The array of search terms (Can only be a word). =item *

    N

    number. The number of web pages results to be looked at.

  • F

    number. Words occuring less than the frequencyCutoff would be excluded from results.

  • BC

    number. Bigrams occuring less than the bigramCutoff would be excluded from results.

  • I

    number. The number of iterations that you want the function to search and build cluster on.

  • S

    number. Takes on the values 1,2 or 3 indicating the relatedness measure to be used.

  • SC

    number. Words and Bigrams with relatedness score greater than the scoreCutOff would be excluded from results.

  • trace.

    string. The location where the file containing the retreived information has to be stored.

  • html.

    bool. If set to "true" then the results returned by the algorithm is in HTML format. If not "true", the results are in plain text format.

returns : Returns an html or text version of the results.

__METHOD__->getWordsInPage(searchTerms,N,C,I, NT,BI,trace)

Purpose:Given a set of search terms, this function will retreive the resulting URLs from Google, it will then follow those links, and retrieve the text from there. Once all the text is collected, the function finds the intersecting or co-occurring words between the top N results. This function is basically used by the function Algorithm1.

Valid arguments are :

  • searchTerms

    string. An array which contains each search term (It can only be a word not phrase).

  • N

    number. The number of web pages to be looked at.

  • C

    number. Words occuring less than the frequencyCutoff would be excluded from results.

  • I

    number. The current iteration number.

  • NT

    number. The number of search terms in the initial set.

  • BI

    number. The bigram cutoff.Set to 0 to exclude bigrams.

  • trace.

    string. The location of the trace file.

returns : Returns nothing.

SENTIMENT CLASSIFICATION

    This set of functions deal with sentiment classification. The functions include the PMI-IR, and some other similar functions that try to classify if a given word or phrase is positively or negatively oriented in its sentiment.

__METHOD__->predictSemanticOrientation(rfile,posInf,negInf,trace)

Purpose: This function tries to predict the semantic orientation of a paragraph of text. The semantic orientation of a paragraph is calculated according to the paper "Thumbs Up or Thumbs Down? Semantic Orientation Applied to Unsupervised Classification of Reviews" By Peter Turney. The difference between Peter Turneys implementation of the PMI-IR algorithm and the implementation of the PMI-IR algorithm in Google Hack is small, but very important.

In Peter Turneys implementation, the PMI-IR algorithm uses the search engine Alta Vista. However, in Google-Hack, we are using Google as our search engine. More importantly, AltaVista provides a "near" operator which the original PMI=IRuses, however, Google does not. Hence, we are using the "AND" operator.

Valid arguments are :

  • rfile

    string. The location of the review file

  • posInf.

    string. Positive inference such as excellent

  • negInf.

    string. Negative inference such a poor

  • trace.

    string. The location of the trace file. If a file_name is given, the results are stored in this file

Returns : the PMI measure and the prediction which is 0 or 1.

__METHOD__->predictWordSentiment(infile,posInf,negInf,html,trace)

Purpose:Given an file containing text, this function tries to find the positive and negative words. The formula used to calculate the sentiment of a word is based on the PMI-IR formula given in Peter Turneys paper.

              (hits(word AND "excellent") hits (poor))

         log2 ----------------------------------------

              (hits(word AND "poor") hits (excellent))

For more information refer the paper, "Thumbs Up or Thumbs Down? Semantic Orientation Applied to Unsupervised Classification of Reviews" By Peter Turney.

  • infile

    string. The input file

  • posInf

    string. A positive word such as "Excellent"

  • negInf.

    string. A negative word such as "Bad"

  • html.

    string. Set to "true" if you want the results to be HTML formatted

    trace.

    string. Set to a file if you want the results to be written to the given filename.

returns : Returns an html or text version of the results.

__METHOD__->predictPhraseSentiment(infile,,posInf,negInf,html,trace)

Purpose:Given an file containing text, this function tries to find the positive and negative phrases. The formula used to calculate the sentiment of a phrase is based on the PMI-IR formula given in Peter Turneys paper.

              (hits(phrase AND "excellent") hits (poor))

         log2 ------------------------------------------
     
              (hits(phrase AND "poor") hits (excellent))

For more information refer the paper, "Thumbs Up or Thumbs Down? Semantic Orientation Applied to Unsupervised Classification of Reviews" By Peter Turney.

  • infile

    string. The input file

  • posInf

    string. A positive word such as "Excellent"

  • negInf.

    string. A negative word such as "Bad"

  • html.

    string. Set to "true" if you want the results to be HTML formatted

    trace.

    string. Set to a file if you want the results to be written to the given filename.

returns : Returns an html or text version of the results.

SPELLING SUGGESTION

__METHOD__->phraseSpelling(searchString)

Purpose: This is function is used to retrieve a spelling suggestion from Google

Valid arguments are :

  • searchString

    string. Need to pass the search string, which can be a single word

Returns: Returns suggested spelling if there is one, otherwise returns "No Spelling Suggested":

Use this function to issue queries to Google.

__METHOD__->Search(searchString,num_results)

Purpose: This function is used to query googles

Valid arguments are :

  • searchString

    string. Need to pass the search string, which can be a single word or phrase, maximum ten words

  • num_results

    integer. The number of results you wast to retrieve, default is 10. Maximum is 1000.

Returns: Returns a GoogleHack object containing the search results.

MANIPULATE WEB TEXT

This set of functions deal with retrieving text from the World Wide Web. Basically, the user can use these functions to retrieve sentences, words, or phrases that occur in web pages (In snippets, cached web pages, links etc.

__METHOD__->getSearchSnippetWords(searchString,numResults,trace_file)

Purpose: Given a search word, this function tries to retreive the text surrounding the search word in the retrieved snippets.

Valid arguments are :

  • searchString

    string. The search string which can be a word or phrase

  • numResults

    string. The number of results to be processed from google.

  • trace_file.

    string. The location of the trace file. If a file_name is given, the results are stored in this file

  • proximity

    string. The number of words surrounding the searchString (Not Implemented) yet

returns : Returns an object which contains the parsed information

__METHOD__->getCachedSurroundingWords(searchString,trace_file)

Purpose: Given a search word, this function tries to retreive the text surrounding the search word in the retrieved CACHED Web pages. It basically does the search and passes the search results to the WebService::GoogleHack::Text::getCachedSurroundingWords function.

Valid arguments are :

  • searchString

    string. The search string which can be a word or phrase

  • trace_file.

    string. The location of the trace file. If a file_name is given, the results are stored in this file

returns : Returns a hash with the keys being the words and the values being the frequency of occurence.

__METHOD__->getSearchSnippetSentences(searchString,trace_file)

Purpose: Given a search word, this function tries to retreive the sentences in the snippet.It basically does the search and passes the search results to the WebService::GoogleHack::Text::getSnippetSentences function

Valid arguments are :

  • searchString

    string. The search string which can be a word or phrase

  • trace_file.

    string. The location of the trace file. If a file_name is given, the results are stored in this file

returns : Returns an array of strings.

__METHOD__->getCachedSurroundingSentences(searchString,trace_file)

Purpose: Given a search word, this function tries to retreive the sentences in the cached web page.

Valid arguments are :

  • searchString

    string. The search string which can be a word or phrase

  • trace_file.

    string. The location of the trace file. If a file_name is given, the results are stored in this file

returns : Returns a hash which contains the parsed sentences as values and the key being the web URL.

__METHOD__->getSearchCommonWords(searchString1,searchString2,trace_file,stemmer)

Purpose:Given two search words, this function tries to retreive the common text/words surrounding the search strings in the retrieved snippets.

Valid arguments are :

  • searchString1

    string. The search string which can be a word or phrase

  • searchString2

    string. The search string which can be a word or phrase

  • trace_file.

    string. The location of the trace file. If a file_name is given, the results are stored in this file

  • stemmer.

    bool. Porter Stemmer on or off.

returns : Returns a hash which contains the intersecting words.

__METHOD__->getWordClustersInSnippets(searchString1,iterations,number,trace_file)

Purpose:Given a search string, this function retreive the top frequency words , and does a search on those words, and builds a list of words that can be regarded as a cluster of related words.

Valid arguments are :

  • searchString1

    string. The search string which can be a word or phrase

  • *=item *

    iterations

    number. The number of iterations that you want the function to search and build cluster on.

    trace_file.

    string. The location of the trace file. If a file_name is given, the results are stored in this file

returns : Returns a set of words as a hash.

__METHOD__->getClustersInSnippets(searchString1,searchString2,iterations,number,trace_file)

Purpose:Given two search strings, this function retreive the snippets for each string, and then finds the intersection of words, and then repeats the search with the intersection of words.

Valid arguments are :

  • searchString1

    string. The search string which can be a word or phrase

  • searchString2

    string. The search string which can be a word or phrase

  • iterations

    number. The number of iterations that you want the function to search and build cluster on.

  • trace_file.

    string. The location of the trace file. If a file_name is given, the results are stored in this file

returns : Returns a hash which contains the intersecting words as keys and the values being the frequency of occurence.

__METHOD__->getText(searchString,iterations,number,path_to_data_directory)

Purpose:Given a search string, this function will retreive the resulting URLs from Google, follow those links, and retrieve the text from there. The function will then clean up the text and store it in a file along with the URL, Date and time of retrieval.The file will be stored under the name of the search string.

Valid arguments are :

  • searchString

    string. The search string which can be a word or phrase.

  • iterations

    number. The number of iterations that you want the function to search and build cluster on.

  • path_to_data_directory.

    string. The location where the file containing the retrived information has to be stored.

returns : Returns nothing.

AUTHOR

Pratheepan Raveendranathan, <rave0029@d.umn.edu>

Ted Pedersen, <tpederse@d.umn.edu>

BUGS

SEE ALSO

WebService::GoogleHack home page - http://google-hack.sourceforge.net

Pratheepan Raveendranathan - http://www.d.umn.edu/~rave0029/research

Ted Pedersen - www.d.umn.edu./~tpederse

Google-Hack Maling List <google-hack-users@lists.sourceforge.net>

COPYRIGHT AND LICENSE

Copyright (c) 2005 by Pratheepan Raveendranathan, Ted Pedersen

This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 2 of the License, or (at your option) any later version.

This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.

You should have received a copy of the GNU General Public License along with this program; if not, write to

The Free Software Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.

1 POD Error

The following errors were encountered while parsing the POD:

Around line 813:

Expected '=item *'