/**
* @dataProvider textByLanguageProvider
*/
public function testByLanguage($languageCode, $text, $expected)
{
$sanitizerFactory = new SanitizerFactory();
$sanitizer = $sanitizerFactory->newSanitizer($text);
$sanitizer->toLowercase();
$text = $sanitizer->sanitizeWith($sanitizerFactory->newGenericRegExTokenizer(), $sanitizerFactory->newCdbStopwordAnalyzer($languageCode), $sanitizerFactory->newNullSynonymizer());
$this->assertEquals($expected, $text);
}
/**
* @dataProvider textProvider
*/
public function testNgramWithBeginEndMarker($languageCode, $ngramSize, $text, $expected)
{
$sanitizerFactory = new SanitizerFactory();
$tokenizer = $sanitizerFactory->newNGramTokenizer($sanitizerFactory->newGenericRegExTokenizer());
$tokenizer->withMarker(true);
$tokenizer->setNgramSize($ngramSize);
$tokens = $tokenizer->tokenize($text);
$this->assertEquals($expected, $tokens);
}
private function predictLanguage($text)
{
if ($this->languageDetection === array()) {
return null;
}
$languageDetector = $this->sanitizerFactory->newNullLanguageDetector();
if (isset($this->languageDetection['TextCatLanguageDetector'])) {
$languageDetector = $this->sanitizerFactory->newTextCatLanguageDetector();
$languageDetector->setLanguageCandidates($this->languageDetection['TextCatLanguageDetector']);
}
return $languageDetector->detect(Normalizer::reduceLengthTo($text, 200));
}
/**
* @dataProvider tinyTextProvider
*/
public function testJaTinySegmenterTokenizer($text, $expected)
{
$sanitizerFactory = new SanitizerFactory();
$tokenier = $sanitizerFactory->newJaTinySegmenterTokenizer($sanitizerFactory->newPunctuationRegExTokenizer());
$this->assertEquals($expected, $tokenier->tokenize($text));
}
