Amharic Normalizer

Normalize Amharic text by mapping phonetically equivalent characters (homophones) to a canonical form.

Why?

Amharic script has multiple characters that represent the same sound. For example:

| Sound | Characters | Normalized | |-------|------------|------------| | /h/ | ሀ ሐ ኀ | ሀ | | /s/ | ሰ ሠ | ሰ | | /ʔ/ | አ ዐ | አ | | /ts'/ | ጸ ፀ | ጸ |

This causes problems when:

• Searching text - "ሰላም" won't match "ሠላም"
• Comparing strings - Two identical-sounding words appear different
• NLP preprocessing - Models treat homophones as different tokens
• Building games/quizzes - User input might use different character variants

This package solves these problems by normalizing all homophones to a single canonical form.

Installation

npm install amharic-normalizer

Usage

import { normalizeAmharic } from "amharic-normalizer";

// Normalize text
normalizeAmharic("ሠላም");  // → "ሰላም"
normalizeAmharic("ሐበሻ");  // → "ሀበሻ"
normalizeAmharic("ዐማርኛ"); // → "አማርኛ"

// Mixed text works too
normalizeAmharic("Hello ሠላም!"); // → "Hello ሰላም!"

// Compare normalized strings
const input1 = normalizeAmharic("ሰላም");
const input2 = normalizeAmharic("ሠላም");
console.log(input1 === input2); // true

Access the Normalization Map

import { AMHARIC_NORMALIZATION_MAP } from "amharic-normalizer";

// Check if a character has a normalized form
console.log(AMHARIC_NORMALIZATION_MAP["ሠ"]); // "ሰ"
console.log(AMHARIC_NORMALIZATION_MAP["ሰ"]); // "ሰ"

Supported Character Families

| Family | Variants | Canonical | Description | |--------|----------|-----------|-------------| | H | ሀ ሐ ኀ (+ all vowel forms) | ሀ series | Glottal fricative /h/ | | S | ሰ ሠ (+ all vowel forms) | ሰ series | Voiceless alveolar fricative /s/ | | A | አ ዐ (+ all vowel forms) | አ series | Glottal stop /ʔ/ | | TS | ጸ ፀ (+ all vowel forms) | ጸ series | Ejective alveolar affricate /ts'/ |

Each family includes all 7 vowel forms (ə, u, i, a, e, ɨ, o).

API

normalizeAmharic(text: string): string

Normalizes Amharic text by replacing phonetically equivalent characters with their canonical form.

Parameters:

• text - The Amharic text to normalize

Returns:

• The normalized text

AMHARIC_NORMALIZATION_MAP: Record<string, string>

A map of Amharic characters to their canonical forms. Characters not in the map are left unchanged.

License

MIT

Author

Eyasu Lingerih