Multilingual Sentiment Analysis

A high-performance multilingual sentiment analysis module for Node.js, leveraging state-of-the-art transformer models (via Hugging Face Transformers) to provide accurate sentiment predictions across multiple languages.

Features

• Multilingual Support: Works out-of-the-box with English, Spanish, French, German, Italian, Portuguese, and more.
• Transformer-based: Uses the distilbert-base-multilingual-cased-sentiments-student model for efficient and accurate inference.
• Simple API: Provides an easy-to-use async interface for analyzing text.
• Reliability Scoring: Includes a confidence score and a reliability flag for each prediction.
• Lightweight Inference: Built with @huggingface/transformers for optimized execution in Node.js environments.

Installation

npm install

Usage

Basic Example

import { getAnalyzer } from 'multilingual-sentiment-analysis';

async function main() {
    const analyzer = getAnalyzer();
    
    // The first call might take a moment to load the model
    const result = await analyzer.analyze('This product is absolutely wonderful!');
    
    console.log(result);
    // { sentiment: 'positive', confidence: 0.99, isReliable: true }
}

main();

Multilingual Support

const esResult = await analyzer.analyze('¡Este producto es excelente!');
// { sentiment: 'positive', confidence: 0.98, isReliable: true }

const frResult = await analyzer.analyze('C\'est terrible, je déteste ça.');
// { sentiment: 'negative', confidence: 0.97, isReliable: true }

Advanced Usage (Custom Models)

You can instantiate the SentimentAnalyzer with a custom model from the Hugging Face Hub:

import { SentimentAnalyzer } from 'multilingual-sentiment-analysis';

const customAnalyzer = new SentimentAnalyzer('Xenova/bert-base-multilingual-uncased-sentiment');
await customAnalyzer.waitForReady();
const result = await customAnalyzer.analyze('I love this!');

API Reference

getAnalyzer(): SentimentAnalyzer

Returns a singleton instance of the default sentiment analyzer.

SentimentAnalyzer Class

• constructor(modelName?: string): Create a new analyzer with an optional custom model.
• analyze(text: string): Promise<SentimentResult>: Analyzes the sentiment of the input text.
• waitForReady(): Promise<void>: Ensures the underlying transformer model is loaded and ready.

SentimentResult Object

• sentiment: 'positive' | 'negative' | 'neutral'
• confidence: number (0.0 to 1.0)
• isReliable: boolean (True if confidence > 0.6)

Development Scripts

• npm run build: Compile TypeScript to JavaScript.
• npm run test: Run the test suite using Mocha.
• npm run evaluate: Run an evaluation script with a battery of test cases across different languages.
• npm run evaluate -- -i: Run evaluation in interactive mode for custom testing.
• npm run lint: Run ESLint to ensure code quality.

License

ISC