Hacker News Reader

Comment by RangerScience | Hacker News Reader

I want a discovery service that shows me the things "most different (but also quality)" from the things I've already seen. It seems like this should be doable.

View on HN

That's trivial to do without using any machine learning.

Go to Goodreads/Librarything, choose a subject/genre you never read and choose the 6th highest rated book (to avoid spam/flash popularity problems)

Personally I'm pretty doubtful this method will give you something you actually want to read though.

> Personally I'm pretty doubtful this method will give you something you actually want to read though.

I've read about 20 books per year for the last three years. I buy most of them through Amazon and always mark them as read on Goodreads in order to get better recommendations.

And yet, I can't remember a single book that I have read exclusively based on showing up as recommended on Goodreads/Amazon. It's usually a word of mouth, some researching, and a couple of classics I've never gotten around to from time to time (currently finishing Fahrenheit 451 for example).

At best, the recommendations end up on my to-read list, usually forgotten, and disappear from there too when I do a cleaning of my to-read list because it got cluttered.

In contrast to your experience, I've read over 150 books in the past 2 years and around 50 of them have come from Goodreads/Amazon recommendations.

I assume it depends on the genres, especially with something like Goodreads. I mainly only read a single genre so it must be fairly easy for them to recommend me books.

Yup, I tend to read a lot of genres, going from highly technical ones, to the fiction and science fiction, all the way down to a couple of YA books when I need something really light just to pass my time.

To make the matter even more complicated, I sometimes read books from my own region, not translated into English (which, of course, makes them unavailable on Amazon), so the algorithm would have to count those ones into recommendations too.

I don't think that anyone is investing a lot of effort in solving this gigantic problem. Amazon just wants to sell you more books, it doesn't matter if you read them or not. I don't think that Goodreads has enough of the capacity to make their recommendations actually work for someone as complex as me.

Subscribe to a magazine with a substantial book review section, perhaps even one of the "The X Review of Books" ones. Besides giving you thoughtful reviews (go for some of the books that don't immediately appeal to you, but that the reviewer is really excited about), the reviews are often very interesting discussions of the books' subjects by writers who know their stuff.

I am curious as to how will you define "quality" here? Is it based on wisdom of the crowds or expert reviews or your own personal consumption pattern?

Could you use something like a convolutional deep net -- in it should have similar low-level features to things you have read, but low scores on higher parts of the network?

I mean, if you generated one for each class of people, you'd probably find that there are easier statistical trends, but conceptually.

In fact, now that I think about it a bit more, you likely could use something like the Netflix algorithm and very broadly read people, who overlap on what you have read, then look for books they've read that score lowest on similarity to things you've read based on inverting the scoring mechanism.

It wouldn't be perfect, but with some tuning you could probably make that work as a simpler model than the DNN one.

What you are saying sounds similar to collaborative filtering[0], which afaik is widely used by likes of amazon. The main challenge I find with the collaborative filtering approach is that it gives lot of weight to my recent behavior so it gets boring very quickly. Not to mention the fact that it starts restricting me to narrow genres & inhibiting the discovery of something new & novel.

[0] - https://en.wikipedia.org/wiki/Collaborative_filtering

What about discovery service called 'talking to people'? Book groups in FB, forums, friends and family... Not only do they actually have good advice, but it has a lot of other merits, too.

Reply on news.ycombinator.com