A cross-platform mobile app that transforms book discovery using an on-device TF-IDF recommendation engine, gamified mini-games, and a privacy-first architecture.
| Login | Home | Search |
|---|---|---|
 |
 |
 |
| My Shelf | Profile Stats | Badges |
|---|---|---|
 |
 |
 |
BookRec goes beyond static lists and rating-based suggestions. Instead of matching genre tags or cover popularity, it analyses the prose of book descriptions using NLP to surface books that genuinely match a reader's taste — all without sending any data to an external server.
- Cover Reveal — Test your visual memory and knowledge of iconic book cover art.
- Sentence Decryption — A unique discovery method that challenges you to identify books from prose and writing style, creating a deeper connection to the material.
- On-Device TF-IDF Engine — Uses a custom-built TF-IDF and Cosine Similarity algorithm to match books based on thematic content and plot descriptions.
- Privacy-First — All recommendation logic runs locally on your device. No reading data ever leaves the device, ensuring fast performance and superior data privacy.
- Dribbble-inspired design with a clean blue and white palette.
- Reading stats, badges & achievements system, Top Shelf profile view.
- Custom squircle shapes and smooth transitions for a high-end feel.
A comprehensive Python-based data engineering suite used to clean, validate, and normalise a 10,000-book dataset, ensuring high-quality metadata and imagery.
| Layer | Technology | Purpose |
|---|---|---|
| Mobile UI | Flutter (Dart) |
Cross-platform Android & iOS app |
| Recommendation | TF-IDF · Cosine Similarity |
On-device NLP-based book matching |
| Local Storage | SQFlite · SharedPreferences |
On-device book data & user prefs |
| Auth & Sync | Firebase Auth · Firestore |
User login & real-time data sync |
| State Management | Provider |
Scalable app-wide state |
| Data Pipeline | Python |
Dataset cleaning, filtering & validation |
lib/
├── models/ # Data models for Books, Users, etc.
├── providers/ # State management logic (Provider)
├── screens/ # UI screens (Home, Profile, Minigames, Admin)
├── services/ # Core logic (Recommendation Engine, Firebase, Database)
├── theme/ # Design system tokens and styling
└── widgets/ # Reusable UI components
- RecommenderService — The heart of the app's TF-IDF content-matching logic.
- MinigamesService — Logic for Cover Reveal and Sentence Decryption games.
- DatabaseService — Manages local data persistence and syncing via SQFlite.
| Script | Purpose |
|---|---|
analyze_dataset.py |
Exploratory analysis of the raw book dataset |
clean_descriptions.py |
Cleans and normalises book description text |
deep_clean.py |
Comprehensive deep cleaning of titles and metadata |
fix_encoding.py |
Corrects mojibake and character encoding errors |
filter_english.py |
Filters dataset to English-language books only |
filter_images.py |
Validates and filters books with accessible cover images |
filter_no_description.py |
Removes entries with missing or empty descriptions |
final_clean.py |
Final pass cleanup before database embedding |
inspect_suspicious.py |
Flags suspicious or anomalous entries for review |
validate_dataset.py |
Ensures structural integrity and data completeness |
eval_recommender.py |
Evaluates recommendation quality on sample inputs |