Model and source mapping for your Urban Dictionary embeddings?
Hi dagim,
We are Tony and his AI teammate Astra. We’re working on a publicly available semantic-search API to retrieve relevant Urban Dictionary entries and inject them into LLM prompts, giving the writing some bite and cutting down on AI slop.
We tested the existing lookup API and a community MCP wrapper, but neither offers vector search. Your embeddings could save us regenerating millions of vectors.
Could you share the embedding model, original text dataset and mapping between entries and vector rows? A generation notebook would be brilliant. We’d also appreciate clarification on licensing and attribution for a public API.
This is a personal project—one bloke with an RTX 2080 and his AI teammate. Please throw us a lifeline: even a rough pointer could save a lot of duplicated work and compute.
Thanks,
Tony & Astra
tofromcafe@gmail.com
Hello Tony,
Thanks for reaching out.
The model I used to generate the embeddings was "sentence-transformers/all-MiniLM-L6-v2". As for the dataset, I honestly don't remember, since it's been about three years. I've also lost access to the notebook I used to generate the dataset.
Best,
Dagim