/**
* Returns the top ranked words
* @param int $amount
* @return array
*/
public function getKeywords($amount = 10)
{
// Map words to their normalized version
$originalToNormalizedMap = [];
$isNormalizedWordIncluded = [];
foreach ($this->wordWeight as $word => $weight) {
$normalized = $this->datasetAdapter->getLemmaFromForm($word);
$originalToNormalizedMap[$word] = $normalized;
$isNormalizedWordIncluded[$normalized] = false;
}
// Sort them by weight
$keywords = $this->wordWeight;
arsort($keywords);
$keywords = array_keys($keywords);
$keywords = array_filter($keywords, function ($content) {
$include = true;
if (is_numeric($content)) {
$include = false;
}
$contentFiltered = preg_replace("/[^a-zA-Z0-9]+/", "", $content);
if (strlen($contentFiltered) < 3) {
$include = false;
}
return $include;
});
// Remove keywords that are already included in a different format
foreach ($keywords as $index => $keyword) {
if ($isNormalizedWordIncluded[$originalToNormalizedMap[$keyword]] == false) {
$isNormalizedWordIncluded[$originalToNormalizedMap[$keyword]] = true;
} else {
unset($keywords[$index]);
}
}
$keywords = array_slice($keywords, 0, $amount);
return $keywords;
}
/**
* Returns IDF for the given word form.
* @param $wordForm
* @return float
*/
public function getInverseDocumentFrequency($wordForm)
{
$frequency = $this->datasetAdapter->getWordFrequency($wordForm);
return log($this->datasetAdapter->getExaminedDocumentCount() / $frequency);
}