Decomposer¶

The Decomposer class handles grapheme decomposition for Sinhala and Tamil text. It encapsulates language-specific phonetic rules, enabling character-level transformations needed for NLP tokenization and metric evaluation.

Class Definition¶

class Decomposer

Module: graphemes_plusplus.decomposer

All methods are @classmethod - no instantiation required.

Functions¶

`decompose(text: str) → str`¶

Decomposes Tamil and Sinhala strings into fundamental phonetic sequences. Punctuation and spaces are unaffected.

decompose(text: str) → str

Parameter	Type	Description
`text`	`str`	Input text containing Tamil or Sinhala characters
Returns	`str`	Decomposed phonetic sequence

>>> from graphemes_plusplus import decompose
>>> decompose("கா")
'க்ஆ'  # consonant base + vowel

How Decomposition Works¶

Tamil¶

Each Tamil grapheme is split into a mei (consonant base + pulli) and an uyir (vowel):

Input	→	Mei	Uyir
க	→	க்	அ
கா	→	க்	ஆ
கி	→	க்	இ
கு	→	க்	உ

Sinhala¶

Sinhala graphemes follow a similar pattern with consonant base + hal + vowel:

Input	→	Base	Vowel
ක	→	ක්	අ
කා	→	ක්	ආ
කැ	→	ක්	ඇ

Language Detection¶

The Decomposer automatically detects the script using Unicode ranges:

Script	Unicode Range
Tamil	`U+0B80` – `U+0BFF`
Sinhala	`U+0D80` – `U+0DFF`

Mixed scripts Non-Tamil and non-Sinhala characters (English, punctuation, numbers) pass through unchanged during decomposition.

Class Constants¶

Tamil Constants¶

TAMIL_VOWELS - 12 Tamil vowels (உயிர் எழுத்துக்கள்): அ, ஆ, இ, ஈ, உ, ஊ, எ, ஏ, ஐ, ஒ, ஓ, ஔ
TAMIL_ACCENT_SYMBOLS - 12 Tamil vowel diacritics: (empty), ா, ி, ீ, ு, ூ, ெ, ே, ை, ொ, ோ, ௌ

Sinhala Constants¶

SINHALA_VOWELS - 18 Sinhala vowels: අ, ආ, ඇ, ඈ, ඉ, ඊ, උ, ඌ, ...
SINHALA_ACCENT_SYMBOLS - 18 Sinhala vowel diacritics
ZWJ_CHARS - Zero Width Joiner character combinations specific to Sinhala