0
$\begingroup$

I have a DB with 2 tables connected with a one to many relationship. Let's say one of A is linked to many of B. The tables have both two fields for a date and some text. And both tables get new entries regularly. I want to train a model that, when a user creates an entry on B he gets suggestions from table A to link.

  • Is this just a simple classification?
  • Wouldn't there be too many classes?
  • Which model would fit best for this task?
  • What would be the steps would I take to do this?
$\endgroup$
1
  • $\begingroup$ Sounds more like clustering. $\endgroup$ Feb 21, 2023 at 10:34

1 Answer 1

1
$\begingroup$

If the entries in A are completely fixed, then you may approach this as a classification problem, assuming that you have enough labeled data to train the model. If the labels can change dynamically, then no.

Maybe you can generate and store sentence embeddings of the entries of both A and B. Then, you may either suggest the entries of A with the closest embeddings to the new entry in B or, if the embeddings of A and B are not semantically related, you may find the closest already existing entry from B and suggest its same entries from A.

$\endgroup$

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.