Good day, just a thing for me to understand:
i am trying to give my assistant some long term memory, achieving this with pinecone and proper embeddings made with the embedding model ada002.
one thing i am unsure about right now is the following:
if my background system handles some things like summary-texts in english, and makes embeddings of it, and later the user is referring to a earlier message in german, the system creates a embedding of this asking message and searches for similar embeddings, does this work as flawlessly as it would in plain english? does anyone here have experience with vector similarity in different languages, even though the content may semtantically be similar?
#Embeddings in different languages?
tardy pike