/**
* -- Add function info here --
*/
function getFarthestToken($blogId, $tokens)
{
$bayesianTokens = new BayesianTokens();
$tokens = $bayesianTokens->getBayesianTokensFromArray($blogId, $tokens, false);
$tempArray = array();
foreach ($tokens as $token) {
array_push($tempArray, abs($token->getProb() - 0.5));
}
arsort($tempArray);
$keys = array_keys($tempArray);
$key = $keys[0];
return $tokens[$key];
}
/**
* @private
*/
function _getMostSignificantTokens($blogId, $tokens)
{
$config =& Config::getConfig();
$bayesianFilterInfos = new BayesianFilterInfos();
$bayesianFilterInfo = $bayesianFilterInfos->getBlogBayesianFilterInfo($blogId);
$totalSpam = $bayesianFilterInfo->getTotalSpam();
$totalNonSpam = $bayesianFilterInfo->getTotalNonSpam();
$bayesianTokens = new BayesianTokens();
foreach ($tokens as $token) {
$bayesianTokens->updateOccurrences($blogId, $token, 0, 0, $totalSpam, $totalNonSpam, false);
}
$tokens = $bayesianTokens->getBayesianTokensFromArray($blogId, $tokens);
$tempArray = array();
foreach ($tokens as $token) {
if ($token->isSignificant() && $token->isValid()) {
array_push($tempArray, abs($token->getProb() - 0.5));
}
}
arsort($tempArray);
$significantTokens = array();
$count = 0;
foreach ($tempArray as $key => $value) {
array_push($significantTokens, $tokens[$key]);
$count++;
if ($count == $config->getValue("bayesian_filter_number_significant_tokens")) {
break;
}
}
return $significantTokens;
}
/**
* untrains the filter
*
* @param blogId The blog id
* @param topic The topic of the comment/article that we're using to untrain the filter
* @param text The text of the comment/articles that we're usingn to untrain the filter
* @param userName Name of the user posting this comment/article
* @param userEmail Email address of the user posting this comment/article
* @param userUrl URL of the user posting this comment/article
* @param spam Wether we should unmark these contents as spam or not. The content will be unmarked
* as non-spam by default
* @static
* @see train
*/
function untrain($blogId, $topic, $text, $userName, $userEmail, $userUrl, $spam = false)
{
$tokenizer = new BayesianTokenizer();
$tokensTopic = $tokenizer->addContextMark($tokenizer->tokenize($topic), TOKEN_TOPIC_MARK);
$tokensText = $tokenizer->tokenize($text);
$tokensUserName = $tokenizer->addContextMark($tokenizer->tokenize($userName), TOKEN_USER_NAME_MARK);
$tokensUserEmail = $tokenizer->addContextMark($tokenizer->tokenize($userEmail), TOKEN_USER_EMAIL_MARK);
$tokensUserUrl = $tokenizer->addContextMark($tokenizer->tokenize($userUrl), TOKEN_USER_URL_MARK);
$tokens = array_merge($tokensTopic, $tokensText, $tokensUserName, $tokensUserEmail, $tokensUserUrl);
$bayesianFilterInfos = new BayesianFilterInfos();
$bayesianFilterInfo = $bayesianFilterInfos->getBlogBayesianFilterInfo($blogId);
$totalSpam = $bayesianFilterInfo->getTotalSpam();
$totalNonSpam = $bayesianFilterInfo->getTotalNonSpam();
$bayesianTokens = new BayesianTokens();
if ($spam) {
$bayesianTokens->decSpamOccurrencesFromTokensArray($blogId, $tokens, $totalSpam, $totalNonSpam);
$bayesianFilterInfos->decTotalSpam($bayesianFilterInfo->getId());
} else {
$bayesianTokens->decNonSpamOccurrencesFromTokensArray($blogId, $tokens, $totalSpam, $totalNonSpam);
$bayesianFilterInfos->decTotalNonSpam($bayesianFilterInfo->getId());
}
return true;
}