AI-driven
for AO3
Way to miss the point. Also, we all know here in the fediverse that AI is bad.
A safe space for aspiring writers, avid readers and explorers of all things fictional. Discuss worlds beyond your wildest dreams, share and chat among fellow readers and writers alike. With fanfiction, the sky is the limit and your imagination is limitless
AI-driven
for AO3
Way to miss the point. Also, we all know here in the fediverse that AI is bad.
AI can certainly have uses, and was already used for some things before the craze took off after COVID. Especially when using open source, local LLMs that you trained yourself and use local resources instead of killing the planet in mass data centers...
But really? AI and AO3??? And needing an LLM to find fic recs????
The tag system is literally right there. You can make it so specific that it's almost insane how easy it is (unless you're looking for tags that haven't been registered yet). It's really crazy seeing people begging for a robot to tell them that they'll like this post. It's refreshing to discover on my own. Or just go to other social media sites, search your fandom in tags, and find fanfic recs that way... I've found so many amazing fics by simply just interacting in the fandom online and reading people's work and posts/comments.
People are rightfully protecting their work, no wonder AI stuff isn't working on there. Staff has made more efforts to block bots, and people are making fics log-in only. I think you need to just take the time to get lost and have fun.
Why would a rec engine have anything to do with LLMs?
Also the social media for my fandoms are either kinda dead, toxic (aka Twitter or Instragram), fanfic unfriendly for inexplicable reasons (Reddit, which has the most bizarre obsession with pure canon), or... Discord :/ I appreciate the site's minimalism, but honestly I wish Ao3 would implement more discovery/social features of their own.
tl;dr: You're out of luck, maybe ask for recommendations in AO3 comment sections. Using AI to recommend fics is the antithesis of the community
You could try this Google Colab Notebook code made by this Reddit post, IDK if it works though.
Long Post:
Why would a rec engine have anything to do with LLMs?
You literally asked for an AI tool to recommend you fics. An LLM is just another fancy name for AI.
A large language model (LLM) is a type of machine learning model designed for natural language processing tasks such as language generation. LLMs are language models with many parameters, and are trained with self-supervised learning on a vast amount of text.
The only way it would be able to recommend you fics, is to use data from sources that it's been fed. It needs to be able to process text so it can understand grammar, but also understand what people like/read to then "recommend" you something that it "thinks you would like" which is usually just more data of what other people similarly did (algos are a tad creepy tbh). Often, these sources are not compensated and are just stolen straight from random posts, articles, stories, etc. on the Internet, including AO3 works.
Unless an algorithm went purely off of kudos & history of users, maybe with some added tags for specification, that's about the only way I could think of recommendations without AI/LLM. But you literally put AI in your post. Not to mention AO3 stalking users history is just... no. They know that would kill the service (due to loss of user trust), and people would use the code to just make an alternative (like SquidgeWorld Archive)
Unfortunately, if your fandom is dead, fic recommendations aren't gonna fix the lack of content on AO3
If they're toxic, then block and move on, or just stop checking that community and look on another site, especially Tumblr. People write fic on there and tend to be a tad more open/welcoming due to so many of them also being AO3 lovers.
IDK why you specifically hate Discord, but if it doesn't work for you, don't use it I guess.
The discovery is pretty damn good when clicking tag first and then filtering, especially if you use the search within results feature. It scans tags, authors, summaries, notes, etc based on what you type. Use the Boolean operators like AND, OR, and NOT (-) to help narrow searches, along with * and " ". People have great guides on searching
Look at the bookmarks of authors you like, or readers who read similar pics to see what they're reading now. I have found a few gems that way. Subscribe to authors that you like, see what they make next, or gift, etc.
Once again, I really don't know how they could make it better, besides more frequent tag wrangling...
An LLM is just another fancy name for AI.
Recommendations engines typically use embeddings models or something similar, where they take two blocks of text and spit out a number as a similarity value. That's simplifying a lot, but the gist of it. At the risk of sounding pedantic, they have nothing to do with LLMs (which are much larger autoregressive text prediction models), and don't need to train on a huge text corpus to function. They considerably predate LLMs, though both can be in the category of NLP.
Unfortunately, they've been associated with toxic AI, I guess... And beside that, much of your point stands, as the Ao3 foundation/users probably want to lean into tagging. You are right, it's better anyway; Ao3's tagging system is fantastic when authors use it.
It sucks that they don't give API access anymore, so tinkerers could set up recs if they want, but still.
Unfortunately, if your fandom is dead, fic recommendations aren’t gonna fix the lack of content on AO3
Ironically I'm thinking of a few big fandoms. The problem is they're too big, and Discord feels like a burgeoning black hole with a bunch of people talking over each other. I also don't like how it siloes off information so no one else can ever find it.
Another is that some are... obsessed with canon, for lack of a better term? Like they actively don't like talking about speculative lore or fanfics, or even canon works beyond the first, like its sacriligious.
I'm just veering into complaining about specific fandoms I guess. I never really got into Tumblr, so maybe I should finally poke my head in there, thanks.