I was trying to understand the POS tagging using this code. During the evaluation, when I used model.evaluate(), I am getting the expected accuracy but when I use model.predict and then take argmax to decide the label, the accuracy drops down to 50%. What could be the reason behind this.
I was trying to understand the POS tagging using this code. During the evaluation, when I used model.evaluate(), I am getting the expected accuracy but when I use model.predict and then take argmax to decide the label, the accuracy drops down to 50%. What could be the reason behind this.