News for hackers. Stuff that matters.
Re:I Want to Fix Goodreads by awaxman11 on Saturday September 12, 2020 @10:43PM (#24457792)
Has anybody ever used stack ranking to power book recommendations? Instead of rating books on a scale of 1-5 star you instead rate each book relative to every other book you’ve ever read.

Some other factors that I think could be used to create a high quality book recommendation engine on top of stack ranking: number of books read (the top 5 books from someone who has read 200+ books is a stronger signal than someone who has read 10 books) and release date (give extra weight to older books to reduce recency bias).

What do you think? Am I onto something or just spewing some sat night nonsense?

Re:I Want to Fix Goodreads by achileas on Sunday September 13, 2020 @09:53AM (#24460849)
Are all books comparable linearly like this though? I could see separate stack ranks for different genres (practical books, sci-fi, relationships, whatever), attributes (readability, teachability if applicable, subject matter interest), or more.

I also can't help but think in graphs, so I could also see horizontal-ish linkages between the stacks as well. Each book, then, could have a number of attributes that get an independent stack ranking for each one, leaving you with a constellation score as well as a focused look at what the book is good at or focused on. I could see it getting unwieldy though.

Thinking through this, maybe genre stacks based on some collapse of attribute ranking scores.

Re:I Want to Fix Goodreads by sam_lynx on Sunday September 13, 2020 @02:27AM (#24458850)
I think this could work, although perhaps in combination, not replacement, with the traditional ranking system. The book on the bottom of somebody’s list isn’t necessarily a bad one.