Show HN: Decide your next read from any booklist basis your Goodreads history
bookrecommendations1-hcr77cz3ga-el.a.run.appHello all! I am a fintech product manager inspired by this community to build things.
The motivation to build this came while going through NYT's List 100 Best Books of the 21st Century. I wanted to see which one I could pick as my next read. I realised that I couldn't get through the descriptions of the 100 books, then go through the process of searching Goodreads ratings for each, read people's comments, evaluate and so on. Could there be something faster which would give me an idea of which books I would like?
So I built a MVP personalized book recommender which takes all your Goodreads reading history (thanks to Goodreads export option), creates a model (neural network) and then gives predicted ratings for any new book i.e. a prediction of how much would you rate it from 1 to 5 stars on Goodreads. This is especially useful if you are trying to decide which books to read from a long list of books. The model will give predicted ratings against each book and you can then choose if you would like to read something which has a high predicted rating or push yourselves towards trying something new.
Since this builds a neural network-based predictive model right on the fly, it takes time to build if your reading history is long. I am trying to see how I can improve on that.
I have just added 2 list of books against which you can get predicted scores - Obama's recommendations and Pulitzer list of books. Would love to hear from you if you want me to add another set of books. A feature in the works is extracting all books from a webpage against which you can get predicted book ratings. For ex: On book discussion pages on Hackernews, there are many books mentioned by community members. This model can help you decide which one to read next in a few moments!
How it works? I am fetching book descriptions basis the title and author of book from Google Books API and then embedding the same using GPT-4. Then building a neural network with embeddings of each books descriptions and the ratings I have given in my reading history. I have used Pytorch to build the same.
Would love to hear suggestions, comments, ideas, critiques!
You can choose to reach out to me at: shubham13596@gmail.com or https://www.linkedin.com/in/shubham-gupta-50267a90/
Thanks! Personally I have always decided my next read from the bibliography at the end of any book. I dont need a manipulative algorithm to make decisions for me. I dont need to "track" my reading, I have a small home library. I order my books from Waterstones or Blackwells. I dont like book reviews. It takes away the excitement of reading. Reading a book is an individual experience and everyone has a different view. I dont read, visit or promote "goodreads" because it is a subsidiary of the Amazon monopoly. Independent book stores not Amazon. Thanks for views. I can understand that such a tool might not be relevant for someone with your reading style. However, this is no manipulative algorithm. It is just a tool to help you decide your next read based on your own reading history.