feat: Add Turkish Syllabification Module (#43)

[ada-cinar]

Summary

Implements research-critical Turkish syllabification functionality following linguistic principles of onset-nucleus-coda structure and Turkish phonotactics.

Motivation

Syllabification (hece ayırma) is fundamental for:

• Prosody Research: Stress patterns, rhythm analysis
• Poetry Analysis: Rhyme detection, meter identification
• TTS/Speech Synthesis: Pronunciation modeling
• Morphophonology: Vowel harmony and suffix patterns

Implementation

Core Algorithm

• Onset Maximization: Single consonants go to next syllable (ki.tap, not kit.ap)
• Turkish Structure: (C)(C)V(C)(C) where V is mandatory
• Cluster Handling: Recognizes valid Turkish onset clusters (tr, pr, kr, etc.)

Turkish Syllable Boundary Rules

V.V       → break between vowels    (o-ku, şa-ir)
V.CV      → C starts next syllable  (ki-tap, a-na)
VCC.V     → split between Cs        (mer-ha-ba, an-la-mak)
VCCC.V    → first C as coda         (türk-çe)

API

Basic Usage

from durak import syllabify

syllabify('merhaba')
# ['mer', 'ha', 'ba']

syllabify('kitap', separator='-')
# 'ki-tap'

Advanced Analysis

from durak import Syllabifier

syl = Syllabifier()
info = syl.analyze('anlamak')
# SyllableInfo(
#     word='anlamak',
#     syllables=['an', 'la', 'mak'],
#     count=3,
#     structure=['VC', 'CV', 'CVC']
# )

Features

✅ Onset maximization (linguistically correct)
✅ Turkish consonant cluster recognition
✅ Case preservation (İSTANBUL → İS-TAN-BUL)
✅ Unicode NFC normalization (handles İ/i properly)
✅ Custom separator support
✅ Detailed syllable structure analysis
✅ Integration-ready (works with tokenizer)

Testing

Comprehensive test suite with 70+ test cases:

• Basic syllabification patterns
• Linguistic accuracy (common words)
• Edge cases (vowel-initial, clusters, apostrophes)
• Integration with tokenizer
• Case preservation
• Circumflex vowels (â, î, û)

Examples

# Poetry analysis - filter by syllable count
words = ['ev', 'kitap', 'merhaba', 'bilgisayar']
three_syllable = [w for w in words if Syllabifier().count(w) == 3]
# ['merhaba']

# Integration with tokenizer
from durak import tokenize, syllabify
tokens = tokenize('Kitabı okudum.')
{t: syllabify(t) for t in tokens if t.isalpha()}
# {'Kitabı': ['Ki', 'ta', 'bı'], 'okudum': ['o', 'ku', 'dum']}

Performance

Pure Python implementation for Phase 1. Future optimization:

• Phase 2: Rust-accelerated version (10-50x faster)
• Batch processing support
• Stress detection
• Morpheme-aware syllabification

Related

Enables:

• [Feature] Implement Tiered Hybrid Lemmatizer in Rust #6 Lemmatizer (syllables inform morpheme boundaries)
• [Feature] Implement Fast TF-IDF and Count Vectorizer in Rust #8 TF-IDF (syllable n-grams as features)
• [Documentation] Add PyTorch and HuggingFace Integration Examples #28 PyTorch/HF integration (syllable embeddings)

Closes #43