Hacker News Reader

> Simply adding the author's name to any query will always take you to the book you are looking for. > ... searching for 'holiday heart robayo' does

Well, this already makes Goodreads worse than any local librarian/book seller. I could go up to the counter of my local library and ask for Inferno and the person at the counter would ask "Are you looking for the one by Dante or by Dan Brown?".

> if you were at a bookstore you would always tell the clerk the author's name and not just say "Do you have Holiday Heart?"

I find this disingenuous. Bookstore clerks don't hit a dead end just because of some missing parameter in the question posted to them. If they did they're losing a lot of potential sales. Plus, any decent inventory software could aide them.

There's any number of authentic reasons why a customer might know only partial info about a book and it's a clerk/librarian's job to help them. "Do you have Narnia books?", "Do you have Lion, Witch, and Wardrobe?", "Do you have this fairy tale from C. S. Lewis" are all questions I'd expect a clerk/librarian to be able to answer adequately.

Even worse, quoting from the article:

> the first five suggestions were books that didn't even contain both words in their titles, despite the site having an entry for the book.

What kind of search system ranks books that don't even have the query that high?

(Disclaimer: I tried searching for "holiday heart" on Goodreads while logged out as a matter of due diligence; I did not get these subpar results. While Robayo's book is nowhere in sight, the search results at least have the search query in their title. Reasonable enough.)

> What kind of search system ranks books that don't even have the query that high?

The author is being disingenuous. The search terms aren't in the book title but they are in the name of the series the book is part of. Presumably because a lot of people search for "Jack Reacher" and expect to find something useful.

That's what they are complaining about.

If I run the google query "holiday heart" site:goodreads.com I get the Margarita García Robayo translation on the first page of 10 results.

On the other hand, when I use the search on goodreads itself, the Robayo book - with 298 ratings - is not among the first 20 results.

Ahead of it are books like:

"Home for the Holidays (Heart of the Wolf #28; Silver Town Wolf #9)" - no phrase match, pluralises holidays, splits the query between the book title and series, and has 178 ratings.

"The Pastry Queen Christmas: Big-Hearted Holiday Entertaining, Texas Style" - again no phrase match, heart becomes big-hearted, and it's a subtitle, and the book has 122 ratings

"Holiday of the Heart (Heart, #4)" - finally a book with both words in the title! But no phrase match, and this book has.... 2 ratings.

If the search doesn't support phrase matches, doesn't rank precise matches higher, and doesn't rank more popular books higher - how the hell is it ranking things?

Of course, goodreads is hardly the first site to have a shitty search that causes users to google things instead. The ranking could well be "whatever order elasticsearch produces"

I've complained about Goodreads a lot, but the idea that their search is bad is new to me. (Except in that I can't open each result in the dropdown in a new tab. Very annoying.)

If I type in Inferno, it drops down Dan Brown's, Sylvian Reynard's, and then Dante's. If I search for Narnia...actually it pops up as the first guess when I just type "N".

Turns out there are edge cases where it does poorly, but it's never been a problem for me, and I search Goodreads pretty often.

EDIT: I should clarify, searching by title or author works really well. Searching by additional metadata is missing, and that's very sad -- I end up just browsing lists. Openlibrary search has a much steeper learning curve, but you can search by reading level, or publisher, or anything.

While the article focused on how someone might bring a better algo webapp against goodreads, you've got to agree that Goodreads is badly maintained. The app is atrocious as is the website from the reader's PoV.

The groups are so badly implemented that it takes multiple clicks to do ANYTHING and there is just no "timeline" like thing where you can post on a group. It won't take much time/resources for a social media to act like a 21st century social media right? Especially since it is owned by the company which created AWS?

from the author side:

I have a very common name and I've got an ebook up on Goodreads(I did not add it & I've not, for the life of me figured out how to submit the book)

But because Good reads thinks that because my name is what it is, they merged my profile with someone else with my exact same first name and last name.

I've done all I could to get them to change it, but they didn't bother to respond.

I love Letterboxd's approach to movie recommendation wherein they don't provide any. Users can create arbitrary lists for all kinds of stuff based anything like genre, themes, feel, mood, cast/crew etc. There is probable a high quality list for anything you can think of. I have found some of my favorite movies on the site this way, no algorithms involved.

I wish more site would do this instead of collecting every single details about my life and still recommend be garbage sponsored content (looking at you Youtube).

Goodreads does have that. It is called lists: https://www.goodreads.com/list?ref=nav_brws_lists

Funny enough, this is another obvious feature that they get completely wrong -- lists are publicly editable, with no (apparent) way of restricting adds to the original creator. So inevitably any niche curatorial effort slides back to the lowest common denominator taste over time.

I was going to say this! Letterboxd is the slickest and most fun media discovery platform I’ve seen. It’s so easy to stumble onto interesting recommendations, and the platform is straight-up fun to use. Look at these lists that Letterboxd users put together:

Insane Toni Collette dinner table cinematic universe https://boxd.it/8DO8A

Or look at this list of films directed by women, organized by poster color: https://boxd.it/lITC

Lots of fun stuff like that, but then also lists composed around incredibly specific vibes:

wlw movies that aren't period dramas and don't have sad endings https://boxd.it/8nKBc (The author of that list says “wlw = an umbrella term referring to lesbian, bi, pan, and queer women”; new to me)

Good movies with no sex scenes to watch with your parents https://boxd.it/54Pcw

The lists themselves are a sort of smart commentary.

Letterboxd gets right the one thing that so many recommendation algorithms get wrong: I'm fickle, and I like different stuff at different times, and I also want new things that I didn't even know I liked. The article about StoryGraph mentions this: too often, algorithms assume you only want more of A) what you listened to last time, or B) what's popular (looking at you, Apple Music).

On a separate topic, one thing I'll say about ratings is that while I find good ratings to be worthless – generally any media of even the most middling quality will have rabid fans somewhere, especially if it's serialized (TV shows, book series, Star Wars, etc.), bad ratings are useful to me; if something is rated badly it's generally because it's truly boring and devoid of any value whatsoever.

Anyway, thanks to the top-level poster for the link to Story Graph. I hope it becomes useful, and in the meantime I'll keep pining for someone to disrupt the rest of Amazon’s book empire as well.

> People still complain about Netflix recommendations and they've spent tens of millions of dollars, possibly over a hundred million dollars, on it and are one of the most valuable companies on the planet with one of the best engineering teams on the planet.

This blog post talks about this exact subject: https://apenwarr.ca/log/20190201

> Why do people think algorithmic recommendations are easy? Or even desirable?

They're easier than you think, it's just most of them on major platforms are optimized to maximize your engagement with the platform, rather than your enjoyment of the content.

You can only say the algorithm works when you know the precise aim of the algorithm. Perhaps the "recommendation" engine is designed to dislodge the user from engaging too much with their own assessment abilities and instead have users respond to less guarded social cues setup elsewhere in the sales funnel.

You can push a lot more pulp on unsuspecting people once it's familiar AND seen in a socially proofed popular list.

Why is nobody asking how a recommendation engine is supposed to work if there’s nothing to recommend? Is it possible that even computers recognize 90% of media as garbage?

Recommendation systems are built on k-Nearest Neighbors (KNN) algorithms. Subjective measure of quality is not a factor. A computer can see that two people like the same thing, but also that those two people like similar things each is not aware of, and bridge that awareness gap. It works the same whether the set of options is Star Wars (original), Star Wars (prequels), and/or Star Wars (sequels).