Algorithm 1.

The process of our algorithm description.

1: Input: Dataset $D={\left\{(x_i,y_i)\right\}}_{i=1}^N$; xi is the input sentence; yi is the corresponding label.

2: The load pre-trained model tokenizes a sentence by splitting the sentence into words or subwords and then pads all lists to the same size.

3: Use the distilBert model to train the dataset to obtain the embedding vector.

4: Put the embedding vector into the logistic regression model to classify the dataset.

5: Model evaluation.