Question

I am using Ian Barbers Naive Bayes analysis class to analyze the sentiment of sentences for a school project. I have created my own datasets of positive neutral and negatives. My problem is I have no clue on how to implement the neutrals and get the class to find them. The link below is for the php class I am using

http://phpir.com/bayesian-opinion-mining

Answer 1

Well the Opinion class is already pretty flexible for adding new “sentiment classes”. Just the classify method has implemented the calculation of “prior” static. But it can be easily replaced with a foreach:

private $classes = array('pos', 'neg', 'neutr');
private $classTokCounts = array('pos' => 0, 'neg' => 0, 'neutr' => 0);
private $classDocCounts = array('pos' => 0, 'neg' => 0, 'neutr' => 0);
private $prior = array('pos' => 1.0/3.0, 'neg' => 1.0/3.0, 'neutr' => 1.0/3.0);

public function classify($document) {
    // remove those:
    //$this->prior['pos'] = $this->classDocCounts['pos'] / $this->docCount;
    //$this->prior['neg'] = $this->classDocCounts['neg'] / $this->docCount;
    // add this:
    foreach($this->classes as $class) {
        $this->prior[$class] = $this->classDocCounts[$class] / $this->docCount;
    }

    // the rest is fine