Research on Chinese Short Text Classification Based on Prefix-vector Attention Template and Probabilistic Answer Set

When people tend to communicate in the network, a large amount of text data is generated. These text data have become an important way for people to obtain information. Nowadays, text classification through prompt learning has also become the focus of people’s research. However, the effect of prompt learning is not ideal due to the immobilization of the prompt template and the single label words. In response to the above problems, this paper proposes a method of constructing a prefix vector attention template (PAP) and a probabilistic answer set. The prefix space vector is added in front of the input text sentence as a template for prompt learning, and using the Attention mechanism to connect the template with the input sentence to construct the input of the model sentence. For each label word, an answer space set is constructed, and the probability that the text belongs to this category is obtained in the form of the answer set probability. The experimental results show that the PAP + probabilistic answer set algorithm proposed in this paper is significantly better than other benchmark models in the text classification task, and can achieve better results.