Book.sv is a free AI-powered book recommendation engine that helps you discover your next favorite book by analyzing reading patterns from millions of Goodreads users.
The tool allows you to input up to 64 books you’ve recently read and generates personalized recommendations based on a machine learning model trained on 3 billion Goodreads reviews from 43 million users.
Features
- AI-Powered Recommendation Engine: Uses a modified SASRec model (a transformer decoder operating on book IDs) trained on A100 GPUs with a vocabulary of 540,000 books.
- Goodreads Import: Allows you to load books from your Goodreads “read” shelf.
- Up to 64 Book Inputs: Accepts a large number of book titles to build a reading profile for more accurate recommendations.
- Similar Books: Displays books similar to any title you select.
- User Intersection Tool: Finds Goodreads users who have read the same combination of books, useful for discovering niche titles and readers with similar tastes.
- Fast Search: Powered by Meilisearch for quick book title lookups.
- 100% Free: No subscription fees, paywalls, or premium tiers (the platform runs with a simple donate option).
Use Cases
- Finding Your Next Read: Input 5-10 books you loved recently and get a curated list of recommendations that match your reading style.
- Genre Exploration: Test the waters in a new genre by inputting a few books you know from that category and discovering similar titles that other readers enjoyed.
- Building Reading Lists: Create themed reading lists by analyzing what readers of specific books tend to enjoy next.
- Discovering Overlooked Books: Use the intersection feature to find readers with similar tastes and explore their lesser-known favorites that didn’t meet the main recommendation threshold.
Case Studies
I tested the tool with three technical programming books: “Clean Code” by Robert C. Martin, “Learning Web Design” by Jennifer Niederst Robbins, and “Eloquent JavaScript” by Marijn Haverbeke. The system generated 30 recommendations that showed a good understanding of the technical reading pattern.
The top recommendations included “You Don’t Know JS: Up & Going” by Kyle Simpson, “JavaScript: The Good Parts” by Douglas Crockford, and “The Pragmatic Programmer” by Dave Thomas. What impressed me was the variety (it didn’t just recommend JavaScript books but also included broader software engineering titles like “Domain-Driven Design” and “Building Microservices”).
Interestingly, the list also included some unexpected titles like “The Subtle Art of Not Giving a F*ck” by Mark Manson and “Steve Jobs” by Walter Isaacson. At first this seemed odd, but it makes sense: many developers who read technical books also read personal development and biography titles. This shows the model captures actual reading patterns rather than just matching by topic tags.
The “Similar” button feature worked well too. Clicking it next to “Clean Code” added that book to my selection and refined the recommendations to focus more heavily on software craftsmanship titles. This gave me a more targeted list for my specific interest area.
How To Use It
1. Visit book.sv and type a book title you’ve read recently. The search results appear instantly thanks to the Meilisearch backend. Click on the correct book from the dropdown results to add it to your “Selected Books” section.
2. Repeat this process for at least three books. The system technically accepts just one or two books, but the recommendations improve significantly with more input. I found that 5-10 books give the best balance between effort and accuracy.
3. If you have a Goodreads account, click the “Import Goodreads” button. This opens a dialog asking for your Goodreads profile URL. The system pulls books from your “read” shelf automatically. This saves considerable time if you’ve been tracking your reading on Goodreads for a while.
4. Once you’ve added your books, click the “Get Recommendation” button at the bottom. The system processes your reading history through the AI model and generates a list of recommended books within a few seconds.
5. Explore the “Similar” button next to each recommended book. Clicking this adds that book to your selected books list, which you can use to refine your recommendations based on a specific interest thread. This helps you go deeper into particular subgenres or topics.
6. If you’re looking for more niche recommendations, try the Intersect feature at book.sv/intersect. Input several books that represent a specific interest area and the tool finds all Goodreads users who have read that exact combination. You can then see what else those users have read.
Pros
- Truly Personalized Suggestions: Recommendations based on actual reading patterns rather than simple genre matching.
- Completely Free: No subscription fees or usage limits.
- Respects Privacy: Includes a self-service opt-out page for users who don’t want their data included.
- Fast Performance: Recommendations generate in seconds thanks to optimized inference serving.
Cons
- Popularity Bias: Books with fewer than 100 reads in the training data aren’t included in recommendations.
- Limited to Goodreads Data: Doesn’t incorporate reading history from other platforms.
- No Explanation for Suggestions: The system doesn’t explain why particular books were recommended.
Related Resources
- SASRec Paper: The original research paper explaining the Sequential Self-Attentive Recommendation model used by book.sv.
- gBCE Loss Paper: Technical paper on the loss function used to train the model, useful for understanding the approach.
- Free AI Tools for Books: Discover more free AI tools for book recommendations.
FAQs
Q: Why can’t I find some books in the search?
A: Books must appear at least 100 times in the training data to be included in the search results and recommendations. This filters out indie publications, very recent releases, and extremely niche titles.
Q: How does this compare to Goodreads recommendations?
A: This tool uses a more sophisticated machine learning approach than Goodreads’ current recommendation system. It was specifically built to address complaints about Goodreads recommendation quality after the Amazon acquisition. From testing both systems, book.sv tends to surface more interesting deep-cut recommendations rather than just pushing bestsellers.
Q: Can I save my book list or recommendations?
A: The tool doesn’t currently have accounts or save functionality. Your selected books and recommendations only persist in your browser session. If you want to save your recommendations, take a screenshot or copy the titles to a note-taking app.
Q: What if I only want recommendations based on one book?
A: After adding a single book to your list, click the “Similar” button next to it to get recommendations based solely on that title.
Q: How does the “Intersect” feature work?
A: It uses a technique called Roaring bitmaps to create a compressed map of which users have read which books. When you input multiple books, it finds the users who appear in the bitmaps for all of those books.
