Skip to content
This repository has been archived by the owner on Mar 9, 2023. It is now read-only.

likeminded.pl: also take into how similar other users rate books #31

Open
billy-mosse opened this issue Nov 29, 2019 · 4 comments
Open
Labels
feature New feature or request

Comments

@billy-mosse
Copy link

Hi! I'm super interested in this feature. Would you like to discuss it and implement it together?

Cheers!

@andre-st
Copy link
Owner

andre-st commented Nov 29, 2019

Hi Guillermo, I agree, that would refine the result; not sure about the extent, though.

1. It would then be how similar other users rate an author, because the program now compares the authors, which gives better results than the first version, which had compared only books:

combinations of same books are more rare than combinations of same authors, while latter still satisfies the 'same taste' condition (the assumption with 'same books' is that likeminded people had the same exposure to the exact same books, but that's questionable - and comparing the authors relaxes this assumption

2. You can have different opinions about an author, for example in the field of "exotic" mathematics, but you can still be likeminded insofar as you both are interested in "exotic" mathematics (a minority) and this user might have other books you would agree on, he knows exotic authors. On the other hand, maybe that's not true for fiction authors and your suggestion would noticeably improve the result. The ultimate goal is detecting good sources for new books.
Edit: I could include a factor "exotic author" (number of readers or the like)

I'll give it a try and possibly make it optional. I can not tell you yet when I have time for an implementation (it's just a side project), so it may take a while. But I think the idea is good enough to try it out.

@andre-st andre-st added the feature New feature or request label Nov 29, 2019
@billy-mosse
Copy link
Author

Hi Andre! Thanks for answering. I got this idea for filmaffinity.com, which implements exactly this idea and has proven useful for me - I've been using the site for the last years.

Of course, the fact that works for movies doesn't mean it works for books - we've got many more authors than movies, right?

@andre-st
Copy link
Owner

andre-st commented Dec 2, 2019

Just to make sure:
Did you mean how similar other users rate a book compared to me: he 5 and me 5 or she 1 and me 5.
Or did you mean the similarity among users as a whole group, e.g. 90% readers rated the book or author with 5 stars?

@billy-mosse
Copy link
Author

The first one.

If I want to find my "book soul mate", I'd do the following calculation: get a list of users with whom I have at least X (maybe 20?) authors in common, and compute the average rating they gave for each author. I would compare those ratings against mine, and count the percentage of times their rate is extremely similar to mine.

For example, suppose I rate:

Murakami: 3.73 avg
Cortazar: 4.55 avg
Coelho: 1.12 avg
Rowling: 3.0

a soul mate would probably have something like:

Murakami: 3.44
Cortazar: 4.1
(and other authors)

while the following person would probably not be a soul mate:
Rowling: 4.9
Coelho: 3.0
Murakami: 3.5

@andre-st andre-st added the next release scheduled for the next release label Dec 14, 2019
@andre-st andre-st removed the next release scheduled for the next release label May 3, 2022
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Labels
feature New feature or request
Projects
None yet
Development

No branches or pull requests

2 participants