Clustex Documentation

Clustex is a lightweight text classification package designed to efficiently categorize text based on similarity metrics and learned token weights.

Constructor

new Classifier(classifications: string[], learningRate?: number, threshold?: number)

Creates a new classifier instance.

• Parameters:
  • classifications (string[]): An array of classification labels.
  • learningRate (number, optional): Influences how quickly token weights adjust. Default is 1.
  • threshold (number, optional): The minimum similarity score for tokens to influence classification. Default is 0.8.

Example:

const classifier = new Classifier(["positive", "negative"], 0.5, 0.9);

Methods

classify(text: string) → string

Determines the most probable classification for the provided text.

• Parameters:
  • text (string): The input text to be classified.
• Returns:
  • The classification label with the highest probability.

Example:

const classifier = new Classifier(["positive", "negative"]);
classifier.example("This is amazing!", "positive");
console.log(classifier.classify("Such a wonderful day!")); 
// Output: "positive"

chance(text: string) → Object

Returns an object with classification probabilities for the given text.

• Parameters:
  • text (string): The input text to analyze.
• Returns:
  • An object mapping classification labels to their probabilities.

Example:

const classifier = new Classifier(["spam", "normal"]);
classifier.example("Buy now!", "spam");
console.log(classifier.chance("Limited-time offer!"));
// Output: { spam: 0.85, normal: 0.15 }

example(text: string, classification: string) → Classifier

Trains the classifier with an example sentence and its corresponding classification.

• Parameters:
  • text (string): The example sentence.
  • classification (string): The label associated with the sentence.
• Returns:
  • The classifier instance (this) for method chaining.

Example:

const classifier = new Classifier(["positive", "negative"]);
classifier
  .example("This is fantastic!", "positive")
  .example("I hate this.", "negative");

stable(stableLearning: StableLearning) → Classifier

Applies a new config from a StableLearning instance to the instance.

• Parameters:
  • stableLearning (StableLearning): StableLearning instance.
• Returns:
  • The classifier instance (this) for method chaining.

Example:

const classifier = new Classifier(["positive", "negative"]);
const stable = new Classifier.StableLearning(classifier);
stable
  .example("This is fantastic!", "positive")
  .example("I hate this.", "negative");
classifier.stable(stable);

dataset(name: string, iterations: number = 1) → Classifier

Loads a predefined dataset and trains the classifier using its entries.

• Parameters:
  • name (string): The dataset name.
  • iterations (number, optional): Number of training iterations. Defaults to 1.
• Returns:
  • The classifier instance (this) for method chaining.

Example:

const classifier = new Classifier();
classifier.dataset("news", 3).classify("Breaking: new policy announced");

datasets

A static array listing available dataset names.

Example:

console.log(Classifier.datasets);
// Output: ["news", "spam", "tone", "importance"]

StableLearning

Classifier.StableLearning

A wrapper class that maintains a reversible training dataset and reinforces classifier stability through repeated forward and reverse passes.

Constructor:

new Classifier.StableLearning(classifier: Classifier)

Method:

example(text: string, classification: string) → this

Trains on an example while replaying all past examples (forward and reverse) to reinforce learning.

License

MIT License