spellcure 1.0.1

SpellCure — correction engine for highly scrambled text

SpellCure

SpellCure is a mathematical correction engine for highly scrambled or distorted text, created by Saheban Khan.
Instead of machine learning, SpellCure uses a position-weighted ratio algorithm to compute similarity between distorted tokens and valid dictionary words.

It is designed to recover words even when the input is heavily jumbled — far beyond simple typo correction.

---

✨ Features

• Corrects heavily scrambled words
• Pure math-based algorithm — no ML required
• Supports both a small built-in vocabulary and NLTK’s large word corpus
• Works on sentences, single words, or mixed noisy text
• Fast, deterministic, and easy to integrate in other projects

---

🧠 How SpellCure Works

SpellCure compares each input token against vocabulary words using:

• position-based character similarity
• ratio scoring
• recursive refinement steps
• optional large NLTK vocabulary

This allows SpellCure to recover the correct word even when the text looks like:

• "aplpe" → "apple"
• "ths si a tset" → "this is a test"
• "tihs wrod is msesdy" → "this word is messy"

---

EXAMPLE from spellcure import SpellCure

Use small built-in vocabulary

model = SpellCure(mode="small") print(model.correct("ths si a tset"))

Output: this is a test # Use NLTK's large vocabulary

model = SpellCure(mode="large") print(model.correct("aplpe"))

🚀 Installation

Install the library using pip:

pip install spellcure