Quick start

Analyze a text

Sentiment::analyze() is the one-liner. It uses a shared default analyzer, built once per language.

use Risan\Sentiment\Sentiment;

$result = Sentiment::analyze('VADER is smart, handsome, and funny.');

$result->label;      // Label::Positive
$result->compound;   // 0.8316
$result->positive;   // 0.746
$result->neutral;    // 0.254
$result->negative;   // 0.0

The compound score is the headline number: -1 is as negative as it gets, +1 as positive as it gets. positive, neutral and negative are ratios that add up to about 1. Understanding the scores explains each one.

Check the label

if ($result->isPositive()) {
    // ...
}

$result->isNegative();
$result->isNeutral();

$result->label is a Label enum, so you can also match on it:

use Risan\Sentiment\Label;

$message = match ($result->label) {
    Label::Positive => 'Glad you liked it!',
    Label::Negative => 'Sorry to hear that.',
    Label::Neutral => 'Thanks for the feedback.',
};

Use it as an array or JSON

$result->toArray();
// [
//     'label' => 'positive',
//     'compound' => 0.8316,
//     'positive' => 0.746,
//     'negative' => 0.0,
//     'neutral' => 0.254,
// ]

json_encode($result); // the same shape, as a JSON object

Analyze Indonesian

Pass a Language case, or its string code:

use Risan\Sentiment\Language;
use Risan\Sentiment\Sentiment;

$result = Sentiment::analyze('Filmnya bagus banget!', Language::Indonesian);
$result = Sentiment::analyze('Filmnya bagus banget!', 'id');

$result->label; // Label::Positive

Customize an analyzer

When you need your own words or a different threshold, build an Analyzer. Every with*() method returns a new instance and leaves the original untouched.

use Risan\Sentiment\Analyzer;
use Risan\Sentiment\Language;

$analyzer = (new Analyzer(Language::Indonesian))
    ->withWords(['cuan' => 2.5, 'bapuk' => -2.0])
    ->withThreshold(0.1);

$result = $analyzer->analyze('Investasinya cuan!');

Example scores

These sentences were scored by the library itself:

Where to go next

• Understanding the scores
• Customizing the lexicon
• API reference