dorukkilitcioglu / books2rec Goto Github PK

Heyy ! I have a team of 6 developers, wanting to contribute to an open source project. We find this very interesting and are willing to contribute to it. Just wanted to enquire are you accepting solutions for the project and also whether you have any new open sub problem statement to be taken up as a project.
Thanking you!
Looking for a response!
Hoping to work on this !

.npy files required in the data load aren't generated during runtime

hpfrec uses Poisson matrix factorization for recommendations. We want to see if it works better than our SVD.

We believe that the genre of a recommended book is very important to whether or not a user will like a book. I want to quantify this idea. Here are some questions that can point us in that direction:

• For each user, what is the genre distribution of their books? Are there certain genres they like more?
• What are the genres of the recommended books for said users? Do they align with user genres?
• How does using pure collab filtering differ from hybrid model when it comes to the genre distribution?
• Does recommending books that are more aligned with the genres the user prefers lead to "better" recommendations?

This paper comes into mind: https://arxiv.org/abs/1711.08379

We should add unit tests wherever possible, and set up Travis CI.

We should add a license so other people can also benefit from our code. I'm thinking either GPLv3, a copyleft license that requires all derivative work that is distributed to be open source, NPOSL which explicitly bans for-profit use, or AGPLv3 which also covers derivative works that are not distributed, but used over a network.

I'm leaning towards AGPLv3.

Should explain our purpose, how to use the WebApp, how to run the notebooks, etc

Some of the Goodreads images are missing. We should try to grab the new urls whenever possible.

We need to implement Mean Reciprocal Rank in order to compare our results to literally the only other paper to use Goodbooks-10k. It seems to be very straightforward to implement.

I think it would be useful if we want to create an mobile app or if anyone wants integrate recommendations from books2rec to his/her website.

We want to try and find ways of having information (ratings and book features) on newer books. Must be obtained legally.

Hardcoded file paths wont' work once we serve this website

Very cool project, I was happy to stumble upon it. I've been using Goodreads for many years, and always been disappointed by the recommendations - I don't think I've read a single book on their recommendations!

Just filing this to note that my recommendations included some books I had read.

For example, I was recommended this:
https://www.goodreads.com/book/show/135479.Cat_s_Cradle

Whereas I have read that book and marked it read on Goodreads. My best guess is that books2rec is considering different editions of the same book as different books - this is the version I have shelved:
https://www.goodreads.com/book/show/8699986-cat-s-cradle

Again, partly a guess, but maybe the system should check if the work_id of the book to be recommended matches the work_id of a book the user has read? (http://fastml.com/goodbooks-10k-a-new-dataset-for-book-recommendations/ talks a bit about work_id vs book_id).

In any case, thanks for the project - very cool.

self explanatory

We have lots of potential models floating around. Figure out which we are going to use and how best to combine them.

Once we are done training SVD and Autoencoder model, update WebApp to use these for recommendations

We need to add references for the models we used.