Sindhi NLP Suite (Aadil Nazar)

👤 About the Author

Aadil Nazar is a Data Engineer and Computational Linguistics Researcher . This suite is the result of intensive research into low-resource language digitization, designed to provide the same level of NLP sophistication for Sindhi that exists for major global languages.

🛠 Project Overview

The Sindhi NLP Suite is a high-performance, integrated toolkit for processing the Sindhi language. It moves beyond simple string matching by incorporating Morphological Analysis , Orthographic Confusion Mapping , and Levenshtein Edit Distance algorithms.

Core Capabilities:

1. The Sindhi Lemmatizer (Morphological Engine)

This isn't just a stemmer; it’s a rule-based lemmatizer that understands the grammatical structure of Sindhi.

• Verb Stemming: Handles complex suffixes like ائيندا (future habitual) or يندڙ (habitual participle) to find the base root.
• Noun Pluralization Rules: Automatically reverts plurals (ending in يون, ون, ين) to their singular masculine or feminine forms.
• Linguistic Metadata: Returns POS tags, gender (masculine/feminine), and number (singular/plural) for every analyzed token.
• Synonym Support: Integrated WordNet lookup to provide contextual synonyms.

2. The Sindhi Spellchecker (Logic-Driven)

A morphology-aware spellchecker that reduces "False Misspellings" by cross-referencing with the Lemmatizer.

• Confusion Map System: Uses a custom mapping to handle phonetically similar characters that are frequently swapped in digital typing (e.g., ھ vs ح, س vs ص, ز vs ذ).
• Edit Distance 1 & 2: Implements optimized algorithms to suggest corrections within one or two character changes.
• Normalization: Strips diacritics (Zabar, Zer, Pesh) automatically to ensure matching is based on core orthography.

🚀 Installation

Bash

pip install aadil-nazar-sindhi-nlp

💻 Technical Usage

Advanced Spellchecking with Suggestions

The spellchecker first checks the dictionary, then the lemma, then the confusion map, and finally calculates edit distances.

Python

from aadil_nazar_sindhi_nlp import SindhiSpellchecker

checker = SindhiSpellchecker()

# Test a word with an orthographic confusion error
# Input: 'اصلاح' (with a common character swap)
result = checker.check("اصلاح") 

print(f"Correct: {result['correct']}")
print(f"Suggestions: {result['suggestions']}")

Deep Linguistic Analysis

Use the Lemmatizer to extract the "DNA" of a Sindhi word.

Python

from aadil_nazar_sindhi_nlp import SindhiLemmatizer

lem = SindhiLemmatizer()

# Analyze a plural inflected word: 'ڪتابن'
data = lem.analyze_word("ڪتابن")

print(f"Root: {data['root']}")      # Output: ڪتاب
print(f"Tag: {data['tag']}")        # Output: noun
print(f"Number: {data['number']}")  # Output: plural

📊 Data Engineering & Performance

• O(1) Lookup: Built on Python sets and hash-maps for near-instant validation against massive datasets.
• LRU Caching: Uses @lru_cache for variant generation, making repetitive sentence processing lightning-fast.
• Unicode Standardized: Built to handle the 52-letter Sindhi alphabet and specific UTF-8 character encodings without corruption.

🗺 The Roadmap (Growing Ecosystem)

This package is the foundation. As a researcher, I am actively building and will soon integrate:

• Sindhi POS Tagger: To identify parts of speech in full sentence contexts.
• Named Entity Recognition (NER): For extracting names, dates, and locations.
• Stopword Filtering: For cleaning Sindhi text for Machine Learning models.

⚖ License

Distributed under the MIT License . See LICENSE.txt for details.