I'm trying to figure out the best way to get started on a project.

I have two separate databases, one is a "Template" db and the other is "Content" db. For each table in the Content db, I want to be able to identify which table in the Template db it is most similar to. In terms of similarity, I only care about structure, meaning the table name, column names, and column types.

Because each table can have a different structure, I'm not sure how to set this up.

My current way of thinking is to vectorize the table name, column names, and col types, and then to do a cosine similarity score of each table in Content to each table in Template.

Does that approach make sense? What can I read up on to learn more about how to do this?




Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.