Pular para o conteúdo principal

Quick Start: Multi-class Classifier

This guide walks through a minimal working multi-class classifier using the NaiveBayes engine with in-memory storage.

1. Build the classifier​

use ByJG\TextClassifier\Lexer\ConfigLexer;
use ByJG\TextClassifier\Lexer\StandardLexer;
use ByJG\TextClassifier\NaiveBayes\NaiveBayes;
use ByJG\TextClassifier\NaiveBayes\Storage\Memory;

$nb = new NaiveBayes(
    new Memory(),
    new StandardLexer(new ConfigLexer())
);

2. Train with examples​

Category names are arbitrary strings you define:

$nb->train('PHP is a server-side programming language', 'tech');
$nb->train('Python is used for machine learning and data science', 'tech');
$nb->train('The cat sat on the mat and looked at the dog', 'animals');
$nb->train('Birds fly south during winter migration', 'animals');
$nb->train('The election results were announced on Tuesday', 'politics');
$nb->train('The new tax policy will affect small businesses', 'politics');

3. Classify text​

$result = $nb->classify('programming language learning algorithms');

echo $result->choice;  // 'tech'
echo $result->score;   // e.g. 0.94

// All scores, sorted descending
// $result->scores => ['tech' => 0.94, 'animals' => 0.53, 'politics' => 0.48]

classify() returns null when no categories have been trained yet.

4. Persist to disk (Memory storage)​

// Save the trained model
$storage = new Memory();
// ... train ...
$storage->save('/var/data/model.json');

// Load it back later
$storage2 = new Memory();
$storage2->load('/var/data/model.json');
$nb2 = new NaiveBayes($storage2, new StandardLexer(new ConfigLexer()));

5. Undo a training sample​

$nb->untrain('PHP is a server-side programming language', 'tech');

Result explained​

classify() returns a ClassificationResult with choice (winning category), score (its confidence, 0.0–1.0), and scores (all categories sorted descending). Scores are Robinson-smoothed values — a higher score means stronger evidence. Returns null when no categories have been trained yet.

Next steps​