- Dynamism takes a user query and then scrapes the web for related links. We then run these links through an embedding model to get vector weights, which we pass through a segmentation model to find semantic similarities, before ultimately pushing those into our 768 dimensional vector database.
- From there we run a parametric U-map implementation to map our 768 dimensional latent space into a 3 dimension graph while maintaining 3 dimensional clusters in order to then present this to the user in the front end.
its-all-vectors/dynamism
Folders and files
| Name | Name | Last commit date | ||
|---|---|---|---|---|