Created
January 17, 2012 21:24
-
-
Save tinychaos42/1628965 to your computer and use it in GitHub Desktop.
The clustering algorithm
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
<?php | |
// no argument, process demo json | |
if(!isset($argv[1])) | |
{ | |
$file = file_get_contents('data.json'); | |
} | |
else | |
{ | |
$file = file_get_contents($argv[1]); | |
} | |
$data = json_decode($file); | |
$functionWords = array('always', 'user', 'service', 'clear', 'very', 'body', 'common', 'really', 'havent', 'return','but', 'on', 'with', 'as', 'for', 'in', 'up', 'just', 'few', 'a','all','an','another','any','both','each','either','every', 'she', 'he', 'him', 'was', 'of', 'who', 'and', 'to', 'it', 'or', 'out', 'not', 'is', 'one', 'be', 'has', 'if', 'you', 'her','his','its','my','neither','no','other','our','per','some','that','the','their','these','this','those','whatever','whichever','your', '-', '0', '1','2','3','4','5','6','7','8','9'); | |
$documentStore = array(); | |
$wordFrequencyInDs = array(); | |
$documentTitles = array(); | |
// create word bags | |
echo "Creating word bags...\n"; | |
foreach ($data->articles as $k=>$document) | |
{ | |
$documentWordbagEntry['words'] = createDocumentWordBag($document); | |
// filter stop-words | |
$documentWordbagEntry['words'] = array_diff($documentWordbagEntry['words'], $functionWords); | |
$documentStore[] = $documentWordbagEntry; | |
$documentTitles[$k] = stripslashes($document->title); | |
} | |
// calculate the relevant index numbers | |
echo "Calculating index numbers...\n"; | |
foreach ($documentStore as $k=>$document) | |
{ | |
$termData = array(); | |
foreach ($document['words'] as $term) | |
{ | |
// don't re-do for the same word again | |
if (!array_key_exists($term, $termData)) | |
{ | |
// tf-idf | |
$termData[$term] = termCount($term, $document['words']) * idfCount($term, $documentStore, $wordFrequencyInDs); | |
} | |
$documentStore[$k]['indexes'] = $termData; | |
} | |
} | |
// get the top x in each document | |
echo "Checking top keywords in each document...\n"; | |
foreach ($documentStore as $k=>$document) | |
{ | |
asort($documentStore[$k]['indexes']); | |
unset($documentStore[$k]['words']); | |
$documentStore[$k]['top'] = getTopXTerms($documentStore[$k]['indexes'], 15); | |
unset($documentStore[$k]['indexes']); | |
} | |
// check if there are correlations | |
echo "Checking correlations...\n"; | |
foreach ($documentStore as $k=>$document) | |
{ | |
$documentStore[$k]['related'] = array(); | |
foreach ($document['top'] as $word) | |
{ | |
foreach ($documentStore as $j=>$document2) | |
{ | |
foreach ($document2['top'] as $word2) | |
{ | |
if (strstr($word, $word2) && !in_array($j,$documentStore[$k]['related']) && $k!=$j) | |
{ | |
$documentStore[$k]['related'][] = $j; | |
} | |
} | |
} | |
} | |
} | |
// create the clusters based on the correlations | |
echo "Creating clusters based on the correlations...\n"; | |
$clusters = array(); | |
foreach ($documentStore as $k=>$document) | |
{ | |
$inCluster = documentInCluster($k, $clusters); | |
if ($inCluster===false) | |
{ | |
$clusterEntry = array($k); | |
$clusters[] = $clusterEntry; | |
} | |
foreach ($document['related'] as $j=>$related) | |
{ | |
$relatedInCluster = documentInCluster($related, $clusters); | |
if ($relatedInCluster===false) | |
{ | |
if($inCluster===false) | |
{ | |
$clusters[sizeof($clusters)-1][] = $related; | |
} | |
else | |
{ | |
$clusters[$inCluster][] = $related; | |
} | |
} | |
} | |
} | |
// Swapping document id-s with titles for readability | |
echo "Swapping document id-s with titles for readability...\n"; | |
foreach ($clusters as $cid=>$cluster) | |
{ | |
foreach ($cluster as $id=>$did) | |
{ | |
$clusters[$cid][$id] = $documentTitles[$did]; | |
} | |
} | |
// Output the results | |
foreach ($clusters as $cid=>$cluster) | |
{ | |
echo "Cluster ".($cid+1)." contents:\n"; | |
foreach ($cluster as $id=>$title) | |
{ | |
echo "\t".$title."\n"; | |
} | |
} | |
/** | |
* Check if the document is already in the current cluster set | |
* @param $id | |
* @param $clusters | |
* @return bool|int | |
*/ | |
function documentInCluster($id, $clusters) | |
{ | |
foreach ($clusters as $cid=>$cluster) | |
{ | |
if (in_array($id, $cluster)) | |
{ | |
return $cid; | |
} | |
} | |
return false; | |
} | |
/** | |
* Puts document's title and content field's words into a flat array | |
* | |
* @param $document | |
* @return array | |
*/ | |
function createDocumentWordBag($document) | |
{ | |
$result = array_merge( createAttributeWordBag(stripcslashes($document->content)), createAttributeWordBag(stripslashes($document->title))); | |
return $result; | |
} | |
/** | |
* Removes punctuation and puts words into flat array | |
* @param $attribute | |
* @return array | |
*/ | |
function createAttributeWordBag($attribute) | |
{ | |
$punctuationPattern = array("+",",",".","-","\"","&","!","?",":",";","#","~","=","/","$","£","^","(",")","_","<",">","\r", "\r\n", "\n", "*", "'"); | |
$text = str_replace($punctuationPattern, ' ', strtolower($attribute)); | |
$result = explode(' ',$text); | |
foreach ($result as $k=>$res) | |
{ | |
if($res === '' || strlen($res)<4) | |
{ | |
unset($result[$k]); | |
} | |
} | |
return $result; | |
} | |
/** | |
* Calculate occurrences of a term in array | |
* | |
* @param $term | |
* @param $textArray | |
* | |
* @return array | |
*/ | |
function termCount($term, $textArray) | |
{ | |
$occurrences = array_count_values($textArray); | |
if (isset($occurrences[$term])) | |
{ | |
// calculate relative frequency (long documents are likely contain proportionally more keywords) | |
return $occurrences[$term]/sizeof($textArray); | |
} | |
else | |
{ | |
return 0; | |
} | |
} | |
/** | |
* Calculate the idf score for a term using termInDocumentStore | |
* | |
* @param $term | |
* @param $documentStore | |
* @param &$wordFrequencyInDs | |
* @return float | |
*/ | |
function idfCount($term, $documentStore, &$wordFrequencyInDs) | |
{ | |
if (!array_key_exists($term, $wordFrequencyInDs)) | |
{ | |
$count = termInDocumentStore($term, $documentStore); | |
$wordFrequencyInDs[$term] = $count; | |
} | |
else | |
{ | |
$count = $wordFrequencyInDs[$term]; | |
} | |
return log(abs(sizeof($documentStore)/abs($count))); | |
} | |
/** | |
* Check if the term is in the document store - for the idf calculation | |
* | |
* @param $term | |
* @param $ds | |
* @return int | |
*/ | |
function termInDocumentStore($term, $ds) | |
{ | |
$count = 0; | |
foreach ($ds as $d) | |
{ | |
if (in_array($term, $d['words'])) | |
{ | |
$count++; | |
} | |
} | |
return $count; | |
} | |
/** | |
* Get the top x terms after the tf-idf has been calculated | |
* @param $wordList | |
* @param $x | |
* @return array | |
*/ | |
function getTopXTerms($wordList, $x) | |
{ | |
$size = sizeof($wordList); | |
$sliced = array_slice($wordList, $size-$x); | |
foreach ($sliced as $term=>$value) | |
{ | |
if($value!="–"); | |
$ret[] = $term; | |
} | |
return $ret; | |
} | |
?> |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Dear tinychaos42, very nice script - could you please provide an example "data.json" File? Thank you very much! Best Regards