This is cool, I quickly learned about some companies outside my region that I didn't know existed.
two other things
1. I get slightly > 1 similarity scores for exact matches for my ticker query. EX: AAPL -> 1.0000000000000007
2.It seems like it's matching exactly on ticker name. It would reduce some friction if I could search by company name as well. Then I wouldn't have to lookup the ticker every time I have a query!
It is a great idea. I think the algorithm needs some work though. Probably some guidance to the LLM could pick out attributes you could use to form a “vector” to answer questions as “what sector is this company in?”, “describe their main customer” etc. Would get expensive though perhaps.
I embed descriptions using Open AI's "text-embedding-ada-002" embedding model[0]. Then I use cosign similarity to find the closest embeddings to that of the selected company[1].
This is cool, I quickly learned about some companies outside my region that I didn't know existed.
two other things 1. I get slightly > 1 similarity scores for exact matches for my ticker query. EX: AAPL -> 1.0000000000000007
2.It seems like it's matching exactly on ticker name. It would reduce some friction if I could search by company name as well. Then I wouldn't have to lookup the ticker every time I have a query!
It is a great idea. I think the algorithm needs some work though. Probably some guidance to the LLM could pick out attributes you could use to form a “vector” to answer questions as “what sector is this company in?”, “describe their main customer” etc. Would get expensive though perhaps.
Apparently a 10-K is a detailed financial report for US public companies https://www.investopedia.com/terms/1/10-k.asp
How is the similarity score calculated? Nike and Google have about the same similarity score relative to Apple.
I embed descriptions using Open AI's "text-embedding-ada-002" embedding model[0]. Then I use cosign similarity to find the closest embeddings to that of the selected company[1].
[0]https://platform.openai.com/docs/guides/embeddings/embedding...
[1]https://scikit-learn.org/stable/modules/generated/sklearn.me...