Taru Luojola

Oh how I would love to have a website which could recommend me #books — and whose recommendations would actually work!

I've been using #Goodreads for years, but one of the features I would most love there is one of the worst: recommendations. They just don't work. They're total bullshit, based on commercial ends more than on artistic attributes of books.

I want working recommendations. Do you know where to get them? And algorithmically, so that they won't run out.

Apr 21, 2017 at 10:02pmWeb
Show threadMatija :books:Apr 22, 2017
@Stoori excuse me for answering your question with another question, but why do you prefer algorithmic recommendations to looking for good books based on recommendations and reviews from peer Goodreaders? More and more things are becoming automated, and I personally cherish the possibility to look for good books on my own, on #Goodreads and other places. E.g. I've read a number of great #books that I learned about through some #podcast episode. Just an idea and my 5c.
Show threadTaru LuojolaApr 22, 2017

@moormaan Algorithms and peer recommendations are not mutually exclusive. I already have peers, what I'm missing is a good algorithm to supplement that.

Because, to be honest, I haven't been completely satisfied with what I've got from my peers. I seem to be bad at communicating what I want, because too often I'm appalled at what is recommended to me. Perhaps an algorithm looking my read list and ratings would look them from an angle my current peers are missing.

Show threadMatija :books:Apr 22, 2017
@Stoori I stand corrected - they are not mutually exclusive. Bare in mind though, the way state of the art recommender systems work is that they compare your reading list to reading lists of all other readers, and detect differences between you and those with which you have significant overlap. So, they are similar (like the same stuff), but different (this is what you haven't read, and might like, because similar peers liked it). The difference is "peers" now become all users of that recom eng.
Show threadMatija :books:Apr 22, 2017

@Stoori so, the algorithm effectively does two things:

1) significantly increase the number of peers to take into account for each specific decision, and

2) significantly decrease the amount of each particular peer that is taken into account (i.e. just the star rating, as opposed to review and a meaningful exchange)

I think that's what's putting an upper limit on the quality of book recommendations. It works great for vacuum cleaners, but books require something more. My 5c.

Show threadMatija :books:Apr 22, 2017
@Stoori 2)* I meant to write "the amount of information from each peer"
Show threadMatija :books:Apr 22, 2017

@Stoori furthermore, recommender engines are static over time (given an unchanging input), while you as a human might feel a desire to look for something different at some point, maybe because of a comment you overheard or a desire to learn a new topic.

To summarize my point: I don't think there can be a recommender engine for books that can match the taste of an advanced reader with a well developed taste - we just have to acquire book-seeking skills to match our nuanced and evolving tastes.

Show threadTaru LuojolaApr 22, 2017

@moormaan Number of peers to take into account is good. The book universe is so huge that it's almost impossible to find the pearls without a huge peer group.

There are other aspects than star rating that can be generalised to algorithmic form, no need to limit the engine to that.

And no, the input is not static: 1) I read more all the time, 2) more books are published all the time, 3) recommendations could be made from different subsets at different times.

Show threadMatija :books:Apr 22, 2017
@Stoori agreed, number of peers is good. Also, you might be right about being able to use more features algorithmically (though I cannot tell how difficult a research project good implementation would be). Admittedly, my pace of reading is slower than the pace of book discovery through "manual" means, so I'm reluctant to submit yet another (important) aspect of my life to the whims of the algorithm. Maybe it's inevitable, we'll see. In the meantime, I'm holding the rearguard for serendipity!
Show threadSumomiApr 22, 2017

@Stoori when trying to find new books for my spouse I use a mixture of goodreads' "Readers Also Enjoyed" & Listopia [1] features and I am looking out for people who seems to have a kind of similar taste. Combining this with Amazon recommendations.

Probably the easiest way might be to find a personal librarian who likes to find the perfect fit and know some books... #librarylove

[1] https://www.goodreads.com/list

Show threadcurrently on cybre.spaceApr 22, 2017
@Stoori on librarything they seem to work better, at least for some niches. probably because users can make recommendations there and others can up/downvote them. but sadly librarything can be annoyingly complicated, it took me long to understand some basic functions.
Show threadcurrently on cybre.spaceApr 22, 2017
@Stoori (in some places, automatic and member recommendations are separated, like in the screenshot, in others they are mixed, i think. it is also possible to rate automatic recommendations.) https://computerfairi.es/media/ommZ9UqJQq244Gg7qzs
Show threadTaru LuojolaApr 22, 2017
@maunzikation Oh, I must explore LibraryThing more. Generally there are less reviews and less community action than in Goodreads, so I have used it only to catagolise own library so far.
Show threadcurrently on cybre.spaceApr 22, 2017
@Stoori yes, it can feel a bit lonely there. I tried to keep both in sync for a while, but importing Goodreads data to Librarything didn't always work as I expected, so I stopped at some point. Maybe I should try it the other way around.
Show threadTaru LuojolaApr 22, 2017

@maunzikation Importing data to Goodreads in order to get recommendations doesn't sound a good idea. Goodreads' recommendations are all bonkers.

”Because you read one Finnish books, how about reading all the other Finnish books?” How about no?

”Because you gave this book three stars, how about reading some other mediocre books?” How about recommending good books instead of mediocricities?

So yes, I use Goodreads to see what my friends read, and for that it's a decent tool.

Show threadcurrently on cybre.spaceApr 22, 2017
@Stoori yes, that was how i pictured it, too. using librarything for recommendations, and goodreads for community.
Show threadJessSloanMay 5, 2017
@Stoori I don't know if that's possible really. I mean, taste is such a personal and dynamic thing that it would be exceedingly difficult to mathematically divine which books a person might like based on previously enjoyed reads.
Show threadTaru LuojolaMay 6, 2017
@JessSloan Sure there is room to do better than give random tosses. If a person gives feedback after reading those suggestions, the next round will yield better results, and so on.