Show HN: I scraped 3B Goodreads reviews to train a better recommendation model

(From the site) >If you visit the "intersect" page, you can input multiple books and find the set of users that have read all of those books. This can be useful for finding longer tail books that weren't popular enough to meet the threshold. For instance, if you like reading about the collapse of the Soviet Union, you could put in "Lenin's Tomb" and "Secondhand Time", and see what other books the resultant users have read.

This is how filmaffinity works, which is the best recommendation system I've tried. They have a group of several dozen 'soulmates', which are users with the most similar set of films seen and ratings given; recommendations are other stuff they also liked, and you get direct access to their lists.

>then finding books that are either controversial or possibly ghettoized

Naively, I’d say the surprises are going to be better if you filter more different friends, rather than more controversial books among your friends. As in “find me a person that’s like me only in some ways, tell me what they love”. Long term this method is much better at exposing you to new ideas rather than just finding your cliques holy wars.

I use `fetch` on relative endpoints so that's odd. There shouldn't be any external API calls on my website other than whatever the Cloudflare captcha uses. I also use HTTPS-only in Chrome and did not experience any issues. I just tested Firefox with HTTPS-only on/off and Safari on my phone and I was able to import shelves for multiple users. Are you sure that you do not have any privacy settings on (can you access your shelf in Incognito mode)?

In some sense, it seems to work well, but the results are sort of nothing special and that's not what I'd personally hope for. I put in three books that are unrelated and got results that compare to a standard book store, either from the same series or other meme startup tech bro recommendations that I'd often literally see on the same shelf. I can't say it's not good, because obviously that's how people browse books and that's what you'd get from reviews, which is perhaps why I never consult reviews for anything.

I put in Thinking in Systems and got a bunch of engineering management stuff which I don't care about. Deep work of course gave me all the rich dad poor dad, steve jobs bio, tim ferriswheel crap which shouldn't surprise me at all. Girl with the dragon tattoo gave me the rest of the series.

Thematic similarity + popularity just seems boring, I'd like something that surfaces unusual deep cuts that I wouldn't necessarily find at the book store on the same shelf, but maybe that I could find if I went to a great library and might be out of print, or that I could find on libgen.

With these:

- Thinking In Systems: A Primer

- Paddle to the Amazon: The Ultimate 12,000-Mile Canoe Adventure

- The Elements of Typographic Style

I was kind of hoping to at least get "Grid Systems in Graphic Design" or something, but mostly got Alchemist, Zen', Into the Wild, almost comically mainstream cuts that of course in some cases I've already read or could find in a Cupertino trash can, not that any of them are not worth reading necessarily, but very typical.

An option to surface rarer choices that combine signals from all the books on the list would be neat, like in the above case, the least read real adventure book that somehow touches on the economics of places travelled through with musings about signage or that just happens to use a similar prose that Robert Bringhurst used to make print design theory not dull. Recommendations that only someone with a real sweaty and weird venn diagram of genuine personal deep interests might conjure up, and that a normal person might say "why the hell would I ever read that" but that otherwise amazing books that are just slept on and might never have found a market, or maybe thematically dissimilar+ conceptually similar in aggregate + unpopular. I'd like to be able to input a seed of inspiration that I haven't been able to find the next deeper step in, rather than all the books on how to start a startup in the garage I don't have. If it's James Hoffman's book on brewing coffee at a high level, I wouldn't want another YouTubers book on brewing coffee at a high level, I'd want the Physics of Filter Coffee, or something in an adjacent sphere grid / tree branch that gives me a way to pursue depth AND breadth but not necessarily the same book by someone else, or the same book with different characters. If I've found a seedling or a mushroom, I'd like to explore the root system of that fruiting body, and then at a certain point find a new seedling based on what I've learned so far, or the one video with 50 views that's somehow the best explanation of how to handle back-pressure in highly concurrent systems after I've realized that I don't know shit about concurrency, but not so deep in the stack that I can't bridge the gap; make the series for me.

Granted, my take here might just be an indictment of reviews in general, or at least those sourced from a generic site like goodreads/amazon which is all about popularity and armchair criticism.

Cool work, How much did it cost to train ? Will the source for training be open source ?

See, now this is an excellent use of LLMs (if we're going to be using them at all). Low stakes if it gets shit wrong, but can provide some really useful and surprising answers!

One request, it would be nice to not have to add Goodreads, since I don't use it. I've love to be able to enter a couple of book titles or an author and just get recommendations!