📚 Pretty nervous about this, but here goes. Here's a side project I've been working on for the past few months. It's called Viberary and it's a semantic search engine. It gives you book recommendations based on ✨vibes. ✨ You enter a search query like "funny scifi" and it returns a list of (hopefully!) good recommendations.

https://viberary.pizza/

There is an about page that explains the data, model, etc. It's still pretty early stages but it's been a labor of love for me.

@vicki THAT IS SUCH A GREAT IDEA!! 📘 💕

@acm_redfox 🎉❤️

@vicki This is a wonderful idea 🙂 Would love to contribute in some way … Haven’t read the full GitHub page yet … will do and see if I can contribute in some form.

@vicki Very cool!

Also I must admit to being drawn in by the .pizza TLD 😂

@diazona that was the plan! 😂

@vicki As delightful as it is impressive. And the About page is a destination of its own
https://viberary.pizza/how

@brohrer ❤️❤️

@vicki super cool

@vicki The about page on this is great... I searched a few things and added a couple to my goodreads want list. I searched "beach reads" and was expecting more results like really fun light books to read on the beach and got more beach themed books, but still some interesting things to add to my list.

@danmayer this is great feedback, thanks!

@vicki Great idea - however, there seems to be a need for some improvement (see screenshot below).

@feinschmeckergarten😂 thanks for the screenshot! It’s definitely a… mood at least

@vicki I love the idea! I got some good suggestions for fantasy books that my kid will love.

One concern: I searched for "Jewish nonfiction" and the first result was an incredibly antisemitic book written by a Holocaust denier promoting conspiracy theories about Jews. 😬

@rochelle Ah, thanks for the feedback heads up. :\. I've deleted that one manually, but will think about a strategic way to check the input data, which may take a bit since this is a side project run by one person 😅 . Thanks again and happy the fantasy search worked well.

@vicki Very cool!

It’s got “I Have No Mouth & I Must Scream” on the funny sci-fi list 😅 I can only imagine the data fidelity problems from good reads!

@reconbot in this case it’s not the data but the matching algorithm. The embedding space is vast and mysterious 😂 http://viberary.pizza/how

@feinschmeckergarten @vicki yeah, I got "Brave new world" and "Ready player one" as the 2 top results when searching for "tech utopia"

Interesting result @coopdot
There seems to be a semantic problem @vicki
Then again, if this is machine learning from the web, and there are enough people mixing up dystopia and utopia, it would explain the recommendations.

@feinschmeckergarten @coopdot yep it depends a lot on the input data, which is weighed heavily on the book summary and reviews on goodreads (you can see the data input and model on the how page.) so if a lot of text mentions dystopia and utopia the model will associate them with the book. There are ways to tune and correct for this and it involves a lot of iteration over time.