Given multiple paragraphs, is it possible to classify an entire paragraph while taking into account the surrounding paragraphs?




I am thinking of average pooling the token embeddings from a specific paragraph. This seems uncommon as I wasn't able to find examples of this online.

It is not quite the same as NER, since we know where the paragraph begins and ends. If using sequence classification on the paragraph, it wouldn't take into account the surrounding text.

  • $\begingroup$ A very common example is the BERT model which uses sentences instead of paragraphs. $\endgroup$
    – Valentas
    Apr 1 at 12:04


Your Answer

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

Browse other questions tagged or ask your own question.