@the-horizon-dev/fast-tokenizer NPM

fast-tokenizer

A high-performance TypeScript library for tokenizing text and performing natural language processing tasks—including sentiment analysis—across multiple languages.

Features

✨ Fast & Efficient - Optimized for performance
🎯 Type-Safe - Written in TypeScript with full type definitions
🔧 Zero Dependencies - No external runtime dependencies
🛡️ Robust - Comprehensive error handling and edge case coverage
📦 Lightweight - Small bundle size
🌐 Multi-Language - Support for English, Spanish, and Brazilian Portuguese (with room for more)

Installation

Install via npm:

npm install @the-horizon-dev/fast-tokenizer

Usage

Tokenization

The core tokenizer supports multiple options including case normalization, diacritic removal, and stop word filtering. For example:

import { Tokenizer } from "@the-horizon-dev/fast-tokenizer";

// Create an instance with default options.
const tokenizer = new Tokenizer({ lowercase: true, removeStopWords: true });
const text = "Hello, world! This is a sample text.";
const tokens = tokenizer.tokenize(text, { minLength: 3 });
console.log(tokens);

// Join tokens back to a string.
const joined = tokenizer.join(tokens);
console.log(joined);

Sentiment Analysis

You can perform sentiment analysis on text using the language‑specific modules. The library provides a unified API that leverages language‑dependent dictionaries, negation rules, and (optional) stemming.

Example (English)

import { SentimentAnalyzer } from "@the-horizon-dev/fast-tokenizer";

// Analyze sentiment for English text.
const analyzer = new SentimentAnalyzer();
const result = analyzer.analyze("I absolutely love this product!");
console.log(result);
// {
//   score: 3,
//   comparative: 0.3,
//   tokens: [...],
//   words: [...],
//   positive: [...],
//   negative: [...]
// }

Example (Brazilian Portuguese)

import { SentimentAnalyzer } from "@the-horizon-dev/fast-tokenizer";

// Analyze sentiment for Portuguese text.
const analyzer = new SentimentAnalyzer();
const result = analyzer.analyze("Eu não gosto deste lugar", "pt");
console.log(result);