⌨️ Smart AutoCorrect Keyboard

copilot/upgrade-ai-autocorrect-assistant

AI-powered autocorrect typing assistant — runs 100% in the browser via GitHub Pages. No backend, no account, no internet required after first load.

A production-ready Android custom keyboard and GitHub Pages web app with intelligent autocorrect powered by edit-distance algorithms.

---

🌐 Web App (GitHub Pages)

A fully client-side autocorrect demo that runs directly in any modern browser — no server, no build step required.

Live Demo

After enabling GitHub Pages (see below), your demo will be available at: https://<your-username>.github.io/Smart-AutoCorrect-Keyboard/

Web App Features

• Real-time autocorrect — suggestions appear on every keystroke
• Damerau-Levenshtein edit-distance implemented in pure JavaScript
• Frequency-ranked suggestions — common words win ties
• Runs 100% offline — no network calls after the initial page load
• Responsive card layout — looks great on mobile and desktop

Web App Files

File	Purpose
index.html	Page structure — input box, result displays, accessible markup
style.css	Modern, responsive card UI
script.js	Edit-distance algorithm and autocorrect logic
dictionary.json	English word → frequency map (used by the autocorrect engine)

Running Locally

Simply open index.html in your browser:

# Option 1 – double-click index.html in your file manager

# Option 2 – serve with Python (avoids fetch() restrictions in some browsers)
python -m http.server 8080
# then visit http://localhost:8080

Deploying to GitHub Pages

1. Push the repository to GitHub (all four web-app files must be in the root of the default branch).
2. Go to Settings → Pages in your GitHub repository.
3. Under Source, choose Deploy from a branch.
4. Set the branch to main (or master) and the folder to / (root).
5. Click Save.
6. GitHub will publish the site — the URL appears at the top of the Pages settings page.

ℹ️ Changes pushed to the selected branch are automatically redeployed within a few minutes.

How the Autocorrect Works

1. dictionary.json is fetched asynchronously on page load.
2. The typed word is lower-cased and looked up directly in the dictionary (O(1)).
   If found → "✅ Correct!"; if not → the edit-distance search runs.
3. Every dictionary word whose length difference from the input is within the allowed threshold is scored:
   • Edit distance is computed with the Damerau-Levenshtein algorithm (handles swaps, inserts, deletes, substitutions).
   • Score = frequency − distance × 50 so common words win ties.
4. The highest-scoring candidate within the tolerance window is shown.

Edit-distance tolerance (mirrors the Android Kotlin implementation):

Input length	Max allowed distance
≤ 3	0 (no correction)
4–5	1
6–8	2
≥ 9	3

---

📱 Android App

A production-ready Android custom keyboard with intelligent autocorrect, word suggestions, bigram prediction, and Clean Architecture. main

---

✨ Features

Feature	Details
🔴 Real-time highlighting	Misspelled words are underlined in red as you type
💡 Top 5 suggestions	Ranked by Levenshtein distance + word frequency
🖱️ Click to replace	Tap any suggestion chip to instantly fix the word
📊 Live statistics	Word count, correction count, accuracy percentage
📱 Mobile responsive	Fluid layout that works on every screen size
📲 PWA / Installable	Add to home screen on iOS & Android
✈️ Offline capable	Service Worker caches all assets after first visit
🔒 100% private	All processing is local — nothing leaves your browser

---

🧠 Algorithm — Levenshtein Distance

The autocorrect engine uses the classic Levenshtein edit-distance algorithm to measure how "close" any two words are:

distance(a, b) = minimum number of single-character edits
                 (insertions · deletions · substitutions)
                 required to transform a into b

Tolerance table

Word length	Max allowed distance
≤ 3	0 — no correction
4 – 5	1
6 – 8	2
≥ 9	3

Suggestions are ranked by a combined score:

score = distance − log₂(frequency) × 0.1

Lower score = better suggestion. This ensures that common words rank ahead of rare words at the same edit distance.

Time complexity

Step	Complexity
Single pair comparison	O(m × n) where m, n are word lengths
Dictionary scan	O(D × m × n) — optimised with length pre-filter
Length pre-filter	Skips any dict word whose length differs by > maxDist

---

📁 Project Structure

Smart-AutoCorrect-Keyboard/
├── index.html        ← Single-page app shell
├── style.css         ← Dark-theme responsive styles
├── script.js         ← Levenshtein engine + UI logic
├── words.json        ← 500+ word dictionary
├── manifest.json     ← PWA manifest
├── sw.js             ← Service Worker (offline support)
├── assets/
│   └── icon.png      ← App icon (192 × 192)
└── README.md

---

🚀 GitHub Pages Deployment

One-click setup

1. Fork or clone this repository
2. Go to Settings → Pages
3. Under Source, choose Deploy from a branch
4. Select branch main (or your default branch) and folder / (root)
5. Click Save

Your app will be live at:

https://<your-username>.github.io/Smart-AutoCorrect-Keyboard/

Local development

No build step required — just open index.html in a browser:

# Option 1: Python simple server (recommended — needed for fetch() + SW)
python3 -m http.server 8080
# Then open http://localhost:8080

# Option 2: Node.js
npx serve .

⚠️ Important: Open via http://localhost:… rather than file:// so that fetch('words.json') and the Service Worker work correctly.

---

📸 Screenshots

Desktop view

The main card with real-time highlighted preview and suggestion chips:

┌─────────────────────────────────────────────┐
│  ● ● ●    autocorrect.ai — typing assistant  │
├─────────────────────────────────────────────┤
│  YOUR TEXT                                   │
│  ┌───────────────────────────────────────┐  │
│  │ Teh quik brwon fox jmps over the…     │  │
│  └───────────────────────────────────────┘  │
│  HIGHLIGHTED PREVIEW                         │
│  [Teh] [quik] [brwon] fox [jmps] over the… │
│  SUGGESTIONS                                 │
│  [① the d=1] [② ten d=2] [③ tea d=2]       │
├──────┬──────┬──────┬────────────────────────┤
│  12  │  4   │  67% │  550                   │
│ Words│Fixes │Accur.│ Dict                   │
└──────┴──────┴──────┴────────────────────────┘

---

🛠️ Customising the Dictionary

words.json is a plain JSON array of strings. Add your own domain-specific terms:

{
  "words": [
    "javascript", "typescript", "react", "your-custom-term",
    ...
  ]
}

The engine will deduplicate and sort the list automatically at load time.

---

🤝 Contributing

Pull requests are welcome! Please open an issue first to discuss what you would like to change.

---

📄 Licence

This project is open source and available under the MIT Licence.

---

Made with ❤️ — runs 100% in your browser