如何减少损失，提高文本分类的准确率？

Question

I am using following code to implement Intent Recognition with BERT using Keras and TensorFlow 2

My training dataset has around 250 intents, each intent having ~80 utterances associated to it. So total utterance is 20K

The base code works fine, however when I use my dataset, the accuracy drops to 1%

Is this happening because the number of intents is huge? If not, can you please suggest if optimisation needs to be modified in order to achieve accuracy in the model?

Thanks!

Answer 1

This probably happened due to two reasons

1. The dataset was imbalanced
2. The number of intents were huge

Rohit