public function testTriGram()
{
$tokens = ["one", "two", "three", "four"];
$expected = ["one two three", "two three four"];
$bigrams = NGramFactory::create($tokens, NGramFactory::TRIGRAM);
$this->assertEquals($expected, $bigrams);
}
/**
* Compute the Pointwise Mutual Information on the collocations
* @return array
*/
public function getCollocationsByPmi()
{
$nGramFreqDist = new FreqDist(NGramFactory::create($this->tokens, $this->nGramSize));
$unigramsFreqDist = new FreqDist($this->tokens);
$dataSet = [];
foreach ($nGramFreqDist->getKeys() as $nGramToken) {
$tokens = explode(" ", $nGramToken);
$tally = 1;
foreach ($tokens as $unigramToken) {
$tally *= $unigramsFreqDist->getKeyValuesByWeight()[$unigramToken];
}
// get probabilities of all tokens
$dataSet[$nGramToken] = log($nGramFreqDist->getKeyValuesByWeight()[$nGramToken] / $tally);
}
arsort($dataSet);
return $dataSet;
}
/**
* Get all the possible phrases
* @return array
*/
public function getPhrases()
{
$phrases = [];
for ($index = $this->nGramSize; $index >= 2; $index--) {
$phrases = array_merge($phrases, NGramFactory::create($this->getTokens(), $index));
}
// you cannot use a phrase if it is a substring of a longer phrase
// we must exclude all of the substring phrases
$add = [];
$remove = [];
foreach ($phrases as $phrase) {
if (isset($remove[$phrase])) {
continue;
} elseif (!isset($add[$phrase])) {
$add[$phrase] = true;
// remove the suffix word
$remove[substr($phrase, 0, strrpos($phrase, " "))] = true;
//remove the prefix
$remove[substr($phrase, strpos($phrase, " ") + 1)] = true;
}
}
return array_keys($add);
}
/**
*
* @param array $tokens
* @param string $separator
* @return array
*/
function trigrams(array $tokens, $separator = ' ')
{
return \TextAnalysis\NGrams\NGramFactory::create($tokens, 3, $separator);
}
