Getting Started

@visulima/content-safety is a robust TypeScript library for detecting banned words, profanity, slurs, hate speech, and inappropriate content across multiple languages. It provides accurate, Unicode-aware text analysis with position information for censoring or highlighting.

Key Features

Multi-Language Support

Comprehensive banned word detection across 19 languages:

• European: English, German, Spanish, French, Italian, Dutch, Polish, Portuguese, Swedish, Russian, Irish
• Middle Eastern: Arabic, Persian/Farsi, Turkish, Azerbaijani
• Asian: Japanese, Korean, Chinese, Hindi

Unicode-Aware

Full support for international scripts and complex text:

• CJK Scripts: Japanese, Korean, Chinese (no word boundaries required)
• RTL Scripts: Arabic, Persian, Hebrew
• Complex Scripts: Devanagari (Hindi), Thai, and others
• Unicode Normalization: Consistent matching with NFC normalization

High Performance

Optimized for production use:

• Eager Regex Compilation: Cache built at module load (no first-call lag)
• Single Pass Checking: All languages checked simultaneously
• Efficient Pattern Matching: Longest phrases matched first

Precise Position Tracking

Detailed match information for flexible handling:

• Start and end indices for each match
• Language attribution (which dictionary matched)
• Support for highlighting, censoring, or rejection

Advanced Features

• Case-Insensitive: Matches regardless of capitalization
• Multi-Word Phrases: Detects compound expressions (e.g., "white trash")
• Leet-Speak Detection: Includes common variants (e.g., "b1tch")
• Word Boundary Aware: Prevents false positives in compound words
• Zero Dependencies: Pure TypeScript, works in browser and Node.js

Quick Start

Installation

npm install @visulima/content-safety

yarn add @visulima/content-safety

pnpm add @visulima/content-safety

Basic Usage

import { checkBannedWords } from "@visulima/content-safety";

// Check clean text
const result = checkBannedWords("Hello, how are you today?");
console.log(result.hasBannedWords); // false
console.log(result.matches); // []

// Check text with banned words
const result2 = checkBannedWords("This contains badword");
console.log(result2.hasBannedWords); // true
console.log(result2.matches[0]);
// {
//   word: "badword",
//   startIndex: 14,
//   endIndex: 21,
//   language: "en"
// }

Use Cases

Content Moderation

Filter user-generated content in forums, comments, and chat applications:

import { checkBannedWords } from "@visulima/content-safety";

function moderateContent(userInput: string) {
    const result = checkBannedWords(userInput);

    if (result.hasBannedWords) {
        return {
            allowed: false,
            reason: `Content contains ${result.matches.length} inappropriate word(s)`,
        };
    }

    return { allowed: true };
}

Text Censoring

Automatically censor inappropriate content:

function censorText(text: string): string {
    const result = checkBannedWords(text);

    if (!result.hasBannedWords) {
        return text;
    }

    let censored = text;
    // Process in reverse to maintain indices
    for (const match of result.matches.reverse()) {
        const replacement = "*".repeat(match.word.length);
        censored = censored.slice(0, match.startIndex) + replacement + censored.slice(match.endIndex);
    }

    return censored;
}

Real-Time Validation

Provide instant feedback in forms and text editors:

function validateInput(text: string) {
    const result = checkBannedWords(text);

    return {
        isValid: !result.hasBannedWords,
        matches: result.matches.map((m) => ({
            position: m.startIndex,
            length: m.word.length,
            language: m.language,
        })),
    };
}

Next Steps

Browser and Server Support

Works in all modern environments:

• Node.js 18+
• Browsers (Chrome, Firefox, Safari, Edge)
• Deno
• Bun
• Edge Functions (Vercel, Cloudflare Workers)
• React Native
• Electron