I am going to work a ton this winter to improve topic/genre accuracy, as I think we have more on Mesopotamia but the system is struggling to tag it correctly. We are using an older NLP/ML system that isn't working as well as newer ones. Update coming soon.
Great to hear. Also, a very interesting topic for me since this is close to my masters degree and current job. For a niche topic like ancient mesopotamia I was surprised. Is there a blog post or something about how the site works behind the scenes?
But I am not a developer, so it only goes into things lightly, like our topic system with NLP/ML and other topics. (hoping to be a dev in a few years enough to work on it as well)
Currently, I'm working as a Data Scientist on document retrieval/text matching, and I have a masters degree in a topic close to computer science and linguistics.
> Here are a few, are these interesting or too mundane?
Thanks, I'll have a look. It's quite interesting, even if it's not very technical, since recently I started thinking about building a book trade or selling network for close friends and their friends (invitation based).
> It is python / Django on the backend, nothing crazy I think.
A solid choice, I think.
Do you use something like openlibrary.org as well?
That is a very cool idea! I've been thinking about something like that to help fund the website, which is a network of book trades, and you get credit for a new used book for every trade. Although I was going to charge $10 a year or $1 a book to go 80% toward authors and 20% toward the website.
> Do you use something like openlibrary.org as well?
When I last looked in 2020, the data quality on open library was really bad, so I didn't use them. We created every book recommended in our system manually, as we needed to find high-quality cover images, and it was an easy way to start.
Eventually, we licensed data from the Nielsen API. It has not been a great experience, but it works "ok." I also looked at Ingram and Bokwer. Now things are 90% automated, but we still have to source a high quality cover for each book as the book APIs have such small images.
I am hoping to expand our book DB in 2025 to all books for a lot of the new features. I am going to look at Open Library again then, as it might work better for that limited functionality. I also could have been too pedantic back in 2020 when i looked at them :)
> That is a very cool idea! I've been thinking about something like that to help fund the website, which is a network of book trades, and you get credit for a new used book for every trade. Although I was going to charge $10 a year or $1 a book to go 80% toward authors and 20% toward the website.
Sounds like a nice way to generate some funding, especially as you already have a number of users on your site to start with. I have no idea how much bureaucracy it takes to be able to take 5% of each transaction or so, so I'll probably leave mine completely free - if I ever get the time to implement it, that is.
> When I last looked in 2020, the data quality on open library was really bad, so I didn't use them.
> Now things are 90% automated, but we still have to source a high quality cover for each book as the book APIs have such small images.
Not great, since that would have been a starting point for me. ;) Cool though, that you can invest into a custom database, since you can tailor it to your task and aren't relying on potentially bad data.
> I am hoping to expand our book DB in 2025 to all books for a lot of the new features.
Is there a public roadmap somewhere for the features?
> I also could have been too pedantic back in 2020 when i looked at them :)
Maybe, maybe not. It takes huge effort to go back and improve a system relying on bad data, especially if it's already a certain size. I know from experience, as my company just did that with a large internal catalogue. Not fun.
I chatted with someone who had a platform like this in the past, and they said it worked really well. They ended up shutting it down when a server lost data or something weird. They also made a fair amount of money from links to buy the book new if there were no used copies available to "order."
>Not great, since that would have been a starting point for me. ;) Cool though, that you can invest >into a custom database, since you can tailor it to your task and aren't relying on potentially bad > data.
Legally, you can probably use the Google Books API for a project like this (whereas I couldn't due to their rules). Or, you could also use Open Library since you don't need great data quality, only the title and author to get this running. For me it was some of the other quality issues that made it not worth my time back then.
>Is there a public roadmap somewhere for the features?
I promise I will update it this weekend. I am trying to find a better embed than this text list (behind the scenes, we have a much better system).
Basically, it is:
Roll out a better ad system for our Founding Author Members (as they are heavily funding the website).
Roll out book series pages (and test a notification system for users there)
Ship a massive update to our bookshelf collections of genres, age groups, and topics. This will visually navigate and break down the most loved books of all time, trending, new, and some other cool stuff.
Improve the accuracy of our genre / topic system (as of right now, it is not doing well). And I am working to add themes and tropes into he mix.
Big improvement to book section UX.
Building a DB of all books to power features needing that going forward (going to try to see if I can use Open Library).
Add a monthly "fav read of month" program for readers.
It is a rough list; still testing and thinking on a lot of these.
As I am trying to really do something cool with email and waiting to make sure engagement looks good with 1,000 subscribers before I start evolving and improving it.
> Legally, you can probably use the Google Books API for a project like this (whereas I couldn't due to their rules). Or, you could also use Open Library since you don't need great data quality, only the title and author to get this running. For me it was some of the other quality issues that made it not worth my time back then.
I think, UX wise I'd be in a similar position to you, since 1.) I'd need a short description of the book as well, and 2.) definitely in german, too. So, I'd probably have to create a lot of data myself.
How do I get to my page? The profile section is very spartan. I'm not even sure whether my profile picture was submitted.