1
$\begingroup$

I have a project idea, where I train a bunch of documents on Doc2Vec and then take a novel, input doc, and ideally be able to be told how similar it is to the docs supplied for training as a whole or how well it "fits" with the training docs. Is there a way to do this?

$\endgroup$

1 Answer 1

0
$\begingroup$

There are many possible approaches:

  • using simple similarity measures (e.g. cosine) to compare the new document against every training document.
  • train a binary classifier to distinguish the reference documents from "anything else". Requires negative examples, it's usually difficult to have a representative sample.
  • Use one-class classification, i.e. train a model using only the reference documents. The model tries to represent this class of documents and considers anything else as negative.
  • Could even consider it as a regression problem, i.e. score documents by how similar they are from the reference documents.
$\endgroup$

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.