1
$\begingroup$

I have tried multi-label text classification with BERT.

Here is the sample input: $15.00 hour, customer service, open to industries

One of the labels is Billing_rate and prediction score looks quite good.

Now my question is if I want to extract $15.00 hour basically feature value out of BERT. Can you please suggest what are my next step options?

$\endgroup$
2
$\begingroup$

You'd need to apply a tagger, either a generic NE tagger or a custom-trained one. The tagger works with each token as an instance, so that you can extract a particular sequence of tokens, e.g.:

$15.00     Begin_Billing_rate
hour       Billing_rate
customer   _
service    _
,          _
open       _
to         _
industries _

Of course in order to train a custom tagger you will also have to annotate your data token by token.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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