ai-zero-shot-classifier

ai-zero-shot-classifier

Checkout the demo for a quick start!

🚀 Introduction

ai-zero-shot-classifier is a powerful, flexible JavaScript library designed to perform zero-shot text classification using pre-trained AI embeddings. The library supports multiple providers and models, enabling you to choose the best AI tools for your project, whether it's OpenAI's models or alternative providers like Groq.

🧐 Why ai-zero-shot-classifier?

The Problem

Traditional text classification requires extensive labeled data and retraining models to adapt to new categories. This process can be costly, time-consuming, and impractical when dealing with constantly evolving datasets or dynamic categories.

The Innovation

ai-zero-shot-classifier eliminates the need for labeled datasets by leveraging pre-trained AI embeddings. It allows for dynamic and task-specific labels, enabling real-time classification across various domains without retraining models. It supports multiple providers and their respective models, making it adaptable to diverse use cases.

✨ Features

• Multi-Provider Support: Works with providers like OpenAI and Groq, enabling integration with models such as GPT, Llama, and others.
• Dynamic Labels: Define your labels dynamically for each classification task.
• Multiple Similarity Functions: Supports cosine similarity, dot product, and Euclidean distance for flexible classification needs.
• Batch Processing: Efficiently handles large datasets with customizable batch sizes and concurrency.
• Highly Configurable: Adjustable settings for embeddings, similarity calculations, and more.
• Seamless Integration: Simple API designed for easy use in Node.js and browser environments.

📦 Installation

npm install ai-zero-shot-classifier

or

yarn add ai-zero-shot-classifier

🚀 Usage

Basic Example with classify Function

import { classify } from 'ai-zero-shot-classifier';

const labels = ['Technology', 'Health', 'Finance'];
const data = [
  'Artificial Intelligence is transforming industries.',
  'The stock market has seen unprecedented growth.',
  'Healthcare advancements are improving lives.'
];

classify({ labels, data, config: { similarity: 'cosine' } })
  .then((results) => {
    console.log(results);
  })
  .catch((error) => {
    console.error(error);
  });

Example with ZeroShotClassifier Class

import ZeroShotClassifier from 'ai-zero-shot-classifier';

const labels = ['Technology', 'Health', 'Finance'];
const data = [
  'Artificial Intelligence is transforming industries.',
  'The stock market has seen unprecedented growth.',
  'Healthcare advancements are improving lives.'
];

// Create an instance of the classifier
const classifier = new ZeroShotClassifier({
  provider: 'openai', // Specify the provider
  model: 'text-embedding-3-small', // Specify the model
  apiKey: 'your-api-key', // API key for authentication
  labels, // Provide labels for classification
  dimensions: undefined, // Pass dimensions as a number here to configure vector dimensions
});

(async () => {
  try {
    const results = await classifier.classify(data, {
      similarity: 'cosine', // Choose the similarity metric
    });

    // perform more classification

    console.log('Classification Results:', results);
  } catch (error) {
    console.error('Error during classification:', error);
  }
})();

⚙️ Configuration Options

Option	Description	Default
similarity	Similarity function to use (cosine, dot, euclidean)	cosine
embeddingBatchSizeData	Batch size for data embeddings	50
embeddingBatchSizeLabels	Batch size for label embeddings	50
embeddingConcurrencyData	Concurrency for data embeddings	5
embeddingConcurrencyLabels	Concurrency for label embeddings	5
comparingConcurrencyTop	Concurrency for top-level comparisons	10
comparingConcurrencyBottom	Concurrency for bottom-level comparisons	10

🛠️ Development

Clone the repository:

git clone https://github.com/a-tokyo/ai-zero-shot-classifier.git

Install dependencies:

npm install

Run the development server:

npm start

Run tests:

npm test

Checkout the demo for a quick start!

_{^{Fallback Demo!}}

🤝 Contributing

Contributions are welcome! Feel free to open issues or submit pull requests.