Php Stemmer Laravel Package

Getting Started

Install via Composer:

composer require wamania/php-stemmer

Begin by stemming a single word—ideal for quick validation or unit testing:

use Wamania\PHPStemmer\Stemmer;

$stemmer = new Stemmer();
$stemmer->setLanguage('en'); // or 'fr', 'es', 'it', 'pt', 'nl', 'de', 'ru', etc.
echo $stemmer->stem('running'); // outputs: 'run'

For Laravel, bind a shared instance in a service provider or use a helper:

// app/Providers/AppServiceProvider.php
public function register()
{
    $this->app->singleton(Stemmer::class, fn () => (new Stemmer())->setLanguage(config('app.locale')));
}

Implementation Patterns

• Search indexing: Stem user queries and indexed document fields before storing or comparing—ideal for lightweight full-text search without external services like Elasticsearch:

function stemTokens(array $tokens, Stemmer $stemmer): array
{
    return array_unique(array_map([$stemmer, 'stem'], $tokens));
}

• NLP preprocessing pipeline: Integrate into model observers or service classes for cleaning user-generated content (e.g., forum posts, product reviews) prior to analysis or similarity scoring.
• Laravel Console Tools: Build custom Artisan commands to batch-stem large text exports or audit vocabulary consistency.
• Language-aware stemmers: Detect user language via Accept-Language headers or profile settings to route to the correct stemmer instance—avoid hardcoding setLanguage() globally.

Gotchas and Tips

• Language support isn't universal: The library covers ~10 major European/Slavic languages (en, fr, es, it, pt, nl, de, ru) but lacks aggressive support for Chinese, Arabic, or agglutinative languages. Check Stemmer::getSupportedLanguages() before assuming availability.
• No context awareness: Stemming is purely orthographic—e.g., university → univers, universe → univers. Don’t use for semantic equivalence tasks like synonym expansion.
• Performance note: stemmers are stateful. Reuse a single instance per language (not per call) for best performance. Avoid creating new Stemmer instances inside loops.
• Extend via interfaces: Though minimal, you can swap stemmer backends by implementing Wamania\PHPStemmer\StemmerInterface if you need custom algorithms (e.g., Porter2).
• Debug tip: Use debug_backtrace() or logging to trace stemming mismatches—especially when search results drop unexpectedly after normalization. Compare stemmed vs. raw term frequencies in logs.