A Semantic Matching Method Based on Dynamic Encoding and Dual-attention

Semantic matching is a critical task in the field of natural language processing (NLP). It aims to compare two sentences and identify the relationship between them. A new semantic matching method is proposed in this paper. First, the pre-trained model BERT is used to dynamically encode the input text into the corresponding encoding vectors. Then, the encoding vectors of the two sequences are respectively input to a bidirectional LSTM with attention mechanism in order to obtain the global vectors of two sequences. Finally, the two global vectors are concatenated and input to a linear layer for computing the final semantic result. This paper compares with other representative methods on the Chinese data set LCQMC. The experimental results show that the method proposed in this paper has a significant improvement.