A visual book recommender | Hacker News Reader

A visual book recommender

nathanrooy.github.io

410 pointssquidhunter54 comments

Reminds me of https://www.literature-map.com

Which is a map of all authors in the world sorted by overlap in readership. I found some of my favorite writers by browsing it.

I wonder which approach is better suited to find something that is spot on to my interests.

When I think of my favorite books, they usually are the most popular books of their authors.

Are there any counterexamples, where an author wrote a book that is more profound than their biggest hit but got overlooked for some reason?

Oh man, Literature Map looks really great for finding recs.

That said, I do think book-level might be much more valuable. My first thought for this was Night in the Lonesome October by Roger Zelazny. I haven't read anything else by him yet because my brother informs me his other stuff is entirely different. Looking at Goodreads, I think that qualifies as far from his biggest hit. Is it "more profound?"? Doubtful, but seems likely that you shouldn't group it with his others. I want recommendations based on the book I like, not the author I mostly might-not.

A better example might be how Stephenie Meyer wrote the extremely popular Twilight books, and also The Host which is much less well-known, and better in many respects. Probably qualifies as more profound, too—it's told from the perspective of a parasitic alien. Picture the Yeerks from Animorphs if you read those.)

I wish it was easier to see some of their books, or even copy and search the author from each page.

> Are there any counterexamples, where an author wrote a book that is more profound than their biggest hit but got overlooked for some reason?

Any reply is going to be somewhat subjective, but All Systems Red is not even in my top 5 books by Martha Wells...

Stella Gibbons is basically only known for Cold Comfort Farm, but IMO many of her other things are better, although in a different way. My favourite is Pure Juliet, which is a beautifully gentle portrayal of someone who is on the Autism spectrum, written before Autism was really a thing that people talked about.

Whoa, this is a great resource. Never heard of it before, it looks like I'm finally able to get some proper book recommendations. Absolutely loving it.

Weird though that for certain cases it spells English author names with Cyrillic alphabet. Like for instance when I center the graph on "Stanisław Lem", I can see names like George Orwell or Terry Pratchett spelled in Cyrillic. I wonder why.

I also like https://shepherd.com/. One of its interesting features is that authors list their five favorite books and say why they like them.

Which reminds me of https://fivebooks.com/, where people from a particular field are asked their top five book recommendations for a given theme. The interview format is great, and I've picked up a few recommended books along the way.

> The Best Apocalyptic Fiction, recommended by Elliot Ackerman

> 1. The King James Bible

Lol, I tried it out and put in one of my favorite books, "Liar's Poker" by Michael Lewis and it suggested me a list of "The best novels to help you understand the rich and dysfunctional."

Spot on! Bookmarked!

Thanks for the nice mention :)

I am about a week away from launching genre pages, age pages, and filters for all those things. So on the hard-science-fiction page, you can filter to see books in a variety of fun ways and keep following your curiosity:

Image showing how it works: https://substackcdn.com/image/fetch/f_auto,q_auto:good,fl_pr...

Hit me up at ben@shepherd.com if you want to try the early preview, just tell me your fav genre and ill send you a link.

Id love to have https://same.energy for book contents.

woah, didn't know about this site! it is super cool!

It feels like what Pinterest would be without the annoying bits.

Be careful, it'll become your defacto image search, and then you'll be really disappointed when you realize it hasn't been updated in ages

Really interesting that my favorite sci-fi book, Pandora's Star is in the middle of a void in the center of a large sci-fi cluster.

It explains why I couldn't find anything like it.

Also very explanatory the fact the Tolkien's The Two Towers is right by its side, because I also love that book.

And now I'm already downloading the other "outliers" close to Pandora's Star.

Pandora's Star has a direct sequel: Judas Unchained, and Peter F. Hamilton has other books in that (The Commonwealth) millieu.

Additional authors you might enjoy are Neal Asher, Iain M. Banks, Alistair Reynolds, and Jack McDevitt.

Yeah, I've read the commonwealth.

But nothing like the first book, although the two are actually a single book sold separately, at least for me.

I just read a neighbor of Pandora's Star in the graph (Coyote) and enjoyed it. Note quite like pandora, but fine.

Allen Steele's work isn't really all that similar, though I suppose Coyote and its sequels come closest. The authors I mentioned are much better bets.

Anvaka's YASIV was an extremely strong tool in this space until Amazon discontinued the API it relied on.

https://twitter.com/anvaka https://twitter.com/yasivcom

Thank you. I agree :)

Kind of hacky but I built something similar to apply the page rank algorithm to the authors referenced between books of various topics, here's the result for science:

https://camjohnson26.github.io/author-graph/science/

https://github.com/CamJohnson26/author-graph

Clearly needs a lot of data clean up but still was very helpful for discovering important scientists and their approximate relative impact

The visualizations remind me of those in a paper I co-authored a while ago (2009) visualizing ~2400 scientific journal / ~5.7M full-text articles: "Semantic Journal Mapping for Search Visualization in a Large Scale Article Digital Library" https://nrc-publications.canada.ca/eng/view/accepted/?id=63e...

Cool visualization, but the model needs work. The closest book to "Harry Potter and The Chamber of Secrets" is "The Victorian Lady's Guide to Sex, Marriage, and Manners".

Sounds about right to me!

I have a theory (after having searched for The Diamond Age) that certain very popular books are ironically not going to be close to similar books because they appear on so many varied reading lists.

There's probably a graph theory phenomena that describes what I'm thinking.

Yeah basically what I think you're saying is that in a weighted graph, if there are edge weights which are orders of magnitude larger then the average it throws off certain models. Like nearest neighbor.

Basically just prune the top and bottom %1 of weighted edges to get an appropriate average. Would be my guess for a fix.

Yep this is 100% true (i run shepherd.com).

Within our data books like lord of the rings and harry potter are nearly impossible to map for "books like" because they are connected to so many other things. I am working now to fine tune our model, but it has been an interesting challenge.

IIRC you could do pretty okay in Netflix's recommendation contest by ignoring what an individual likes and just recommending the stuff that "everybody" likes.

Hypothesis: the sort of person who reads "The Victorian...", is likely to both like Harry Potter books, and also to review them enthusiastically online. The typical HP child reader, does not review books online. Just an hypothesis.

> Only include reviews which came from users who had at least 10 reviews.

Not sure if that's a good idea. It shrinks the set of genuine readers and overweights the set of professional spammers.

This looks super cool, but why not use this tool as a non-visual tool to show similar books given a title? As far as I know there aren't many tools for this

Also it would be super cool if we could import out goodreads reading lists and see them on the cluster

This is super cool! I actually have been working on a new book recommendation site (https://braincandy.com) that has a similar-ish (but much smaller scale) visualization for book similarity. It is really interesting how certain genres tend to be much more insular than others and it can be a real challenge to break out of genre boxes when making recommendations. There's so many books out there on the edges and in-betweens that get lost when they don't fit neatly into an existing popular genre, and those indeed can be some of the most interesting.

This is awesome. I only wish the author haden't waited years after scrapping! Many books I've loved have been released in the past couple years

This is an awesome visualization, I am so impressed :). I've been working on (shepherd.com) with a similar goal to try to bring in human groupings to try to determine book connections via the many angles humans bring to the table. And more serendipity like wandering a local bookstore. I really love how you have done this.

Dropping you an email in a few hours :)

Impressive Tool. I would love to have the same for movies.

I'm surprised there are enough books published under the category of "Reverse-harem" to make it its own category along with things like "Horror", "Fantasy", "Business", etc.

Nice article. Not what I was expecting...

I was really hoping this would address "Visual Book" like in ふしぎの海のナディア, Fushigi no Umi no Nadia/The Secret of Blue Water. The LaserDiscs used to say "Visual Book" in every episode.

Great animated series (to me), a mix of 20,000 Leagues Under the Sea and the Illuminati.

Do NOT watch it with the english subs. Suffer with the Japanese, even if You don't speak it. You don't need the junk verbal translation, You will still get the main concepts. And Hanson's driving is so much more manaical in Japanese.

Did I miss how we are supposed to get recommendations from what OP built?

Find something on the map that you know you like. Now look around it to see similar (i.e. if you like this, you're likely to like that) books.

Use the magnifying glass on the visualization to type in a book.

For post-t-SNE processing to get non-overlapping items, see also: https://github.com/Quasimondo/RasterFairy

I also used more crude algorithms that sort by X, group elements in buckets, and within each, sort by Y. Then we get a grid of elements. The result is less high-quality than with iterative algorithms (and depends on if we sort by X or Y first), but it is hard to beat its simplicity.

This is really cool!

I wish the accompanying article was longer. I can't fully grasp how it was done because I don't know enough about the concepts mentioned.

I think that people in general don't read "enough", or maybe can't "enough." By that I mean, recommendations based on other people's reading habits aren't going to be as fine tuned as say, recommendations based on people's music listening habits.

Really cool!

Reminds me a lot of some of the stuff I used to work on at Steam Labs: https://store.steampowered.com/labs

I'm astonished by how well this works. I looked for a bunch of books I'd read recently and kept finding other books I've read or want to read near them.

Love this. Bought a couple of books similar to "Midnight's Children' - author should definitely think about adding affiliate links!

This is great. I think visual tools like this are under utilized. They’re fun to use and can often reveal interesting insights.

similiar t-SNE visualisation, just for papers:

https://static.nomic.ai/pubmed.html

running on their deepscatter visualization engine:

https://github.com/nomic-ai/deepscatter

that keeps things dynamic for rendering

Have you presented this to LibraryThing?

Funny that most remote and isolated clusters ended up being m-m-romance and ... manga.

The gang of four design patterns book being two away from a MAGA book is funny.

Nice! Is there a github repo?

Made an account just to comment on this. Look incredible.

Something's wrong here. I was very excited to explore this, until I searched for "Getting Things Done" by David Allen. A nearby book - "Crippled America: How to Make America Great Again" by Donald J. Trump.

Reply on news.ycombinator.com