NaiveBayes Class

ByJG\TextClassifier\NaiveBayes\NaiveBayes is the multi-class Naive Bayes text classifier.

Constructor

new NaiveBayes(
    StorageInterface  $storage,
    LexerInterface    $lexer,
    ConfigNaiveBayes  $config    = new ConfigNaiveBayes(),
    ?LlmInterface     $llm       = null,
    ?ConfigLlm        $configLlm = null,
)

Parameter	Type	Description
$storage	ByJG\TextClassifier\NaiveBayes\Storage\StorageInterface	Persistence backend
$lexer	ByJG\TextClassifier\Lexer\LexerInterface	Tokeniser
$config	ConfigNaiveBayes	Smoothing parameters (optional)
$llm	LlmInterface|null	Optional LLM for low-confidence escalation
$configLlm	ConfigLlm|null	LLM escalation thresholds (defaults apply when null)

Methods

train()

public function train(string $text, string $category): void

Trains the classifier with $text as an example of $category. The category is created automatically if it does not exist.

• Increments the document count for $category by 1
• Increments the token count for each unique token in $text, scaled by its occurrence count

untrain()

public function untrain(string $text, string $category): void

Reverses a previous train() call. Decrements document and token counts. If a category's document count reaches zero, it is removed from storage.

getCategories()

public function getCategories(): array<string>

Returns the list of categories currently present in storage.

classify()

public function classify(string $text): ?ClassificationResult

Classifies $text and returns a ClassificationResult, or null when no categories have been trained yet.

Return value	Meaning
ClassificationResult	Classification succeeded
null	No categories trained, or only one category exists (one-vs-rest requires ≥ 2)

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

echo $result->choice;   // 'tech'
echo $result->score;    // 0.91

When an LLM is injected and autoLearn=true, the model is retrained on the LLM decision before returning. statScores always reflects the raw score before any LLM involvement.

ClassificationResult fields

Field	Type	Description
choice	string	Winning category name
score	float	Final score of the winning category 0.0–1.0
scores	array<string, float>	All final scores, sorted descending
statScores	array<string, float>	Raw statistical scores before any LLM escalation
llmDecision	string|null	LLM's label if consulted, otherwise null
escalated	bool	true when the LLM was invoked

Usage example

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())
);

$nb->train('PHP is a programming language', 'tech');
$nb->train('The dog ran in the park', 'animals');

$result = $nb->classify('programming language');
echo $result->choice;  // 'tech'
echo $result->score;   // e.g. 0.87

Related

• ConfigNaiveBayes reference
• ConfigLlm reference
• How NaiveBayes works
• Training guide
• Classifying guide
• LLM-Assisted Classification