Calculate cosine simialrity between vector embeddings

[warrencowan]

Hi wiser people

I have 2 tables of embeddings, table 1 is list of terms, and table 2 is a list of categories. Like this

I want to take a row from the terms list, and score its embedding against a list of embeddings in the categories table using a cosine simialrity calculation. I can join the table data together, but dont know how to implement the calc.

Anyone know how to implement that kind of calc in Alteryx?

Many thnx in advance.

w

[OllieClarke · Accepted solution]

Hi @warrencowan 

Not an expert, but based on the formula described here this is how I'd do this:


The first bit is structuring the data as a paired list:


From there its fairly simple :)

Hope that helps,

Ollie

[warrencowan]

thnks @OllieClarke for this detailed overview. let me give it a try. much appreciated.

[warrencowan]

thnx @OllieClarke thats a sweet little solution there.

you're a 🔷

[OllieClarke]

@warrencowan Happy to help :)