Hi everyone, hope you’re all doing well.
Forgive me if this is a silly question - I’m relatively new to the Bubble world.
I am creating a social media platform oriented around written content, with each post containing 500-10,000 words of body content. Users are directed to a feed and presented with a series of posts, initially based on the tags and genres they select during onboarding. Users can also create their own posts, which they label with tags as appropriate.
I’m looking to create the kind of content personalisation/machine learning-driven recommendation system of Tiktok (albeit not as good). Through research, it seems like there isn’t a good way to do this on Bubble, but I was wondering if the following might work:
-
Connect to the OpenAI API, and when a post is generated, have the LLM create a one-sentence summary of the post based on its body content and tags.
-
Store this summary in my database as a field under “Post”.
-
When a user enters the story feed, connect to OpenAI again and ask it to use the respective summaries of the current user’s liked posts to search the various summaries on my database and identify one that a user might like.
-
Once a recommendable post has been identified, retrieve it and push the post out to the user on the social feed.
Later on, might it be possible to have ChatGPT go through each user’s liked posts summaries and create a meta-summary of the user’s interests? Using each user’s meta-summary, could I not then use the same process to find users similar to the current user, and push out more content that similar users have enjoyed (similar to a real recommendation algorithm)?
In essence, I’m trying to get ChatGPT to do the heavy lifting by using its processing and analysis capabilities to create a recommendation system. Even if it works, I know it won’t be nearly as good as Tiktok’s algorithm, but I’m hoping it would be functional.
Does anyone have any idea how to go about this? I know that I need to integrate the OpenAI API, but I’m not sure if the rest is possible or indeed desirable.
Thanks in advance!