This book recommendation system is built completely from scratch: from scraping raw book data, to processing and vectorizing it, and finally serving recommendations through a modern web interface.

The pipeline starts by scraping book information from the source (titles, authors, descriptions, genres and other metadata). The text data is then cleaned, normalized and transformed into numerical vectors using TF‑IDF, so the model can understand the similarity between different books.

On top of this vector space, the system retrieves the most similar books for any given title or query, returning personalized suggestions instantly. The entire flow—from scraping, data processing and vector creation to the front‑end experience—is designed and implemented by me.

Highlights

• Custom web scraper to collect and refresh the book catalog.
• Clean preprocessing pipeline to handle messy real‑world text data.
• TF‑IDF based similarity search for fast, content‑based recommendations.
• Interactive Svelte front end for exploring and filtering suggested books.