# How to classify neutral sentiments using BERT

We can do text classification as positive and negative as mentioned in below notebook. But is there any way to classify neutral sentiment also?

Actually, I want to know like what kind of changes do we need to make in the above notebook so that it can classify neutral sentiments also besides positive and negative.

• The dataset that is used in this notebook has only two labels (look at the polarity column which only has 0 and 1). If your data has more than two labels, I do not see much change needed except minor modifications such as load_dataset function that generates the polarity and label_list array that contains the labels. Other parts seem generic and very generalizable. – Borhan Kazimipour Jun 11 '19 at 7:34