0
$\begingroup$

I am working on the Reddit dataset for sarcasm detection but the sarcastic data points(1) are showing zero percent recall, precision, and accuracy however nonsarcastic are showing 100% recall and 50 precision. Below is the confusion matrix. I don't know why is it not showing any results on sarcastic class.

enter image description here

$\endgroup$

1 Answer 1

0
$\begingroup$

More than precision/recall more concerning is the fact that all your data is being predicted as non-sarcastic. This seems like data cleanliness issue. You need to make sure your train/validation/test data is split and no overlap is present. Also make sure both the classes are present in all the 3 splits.

$\endgroup$

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.