Two-Stage Text Summary Model

In order to solve the problems of redundant information processing and high quality summary generation in existing methods, this paper proposes a two-stage text summary model which is composed of abstracted and generated models. First of all, the important information is abstracted by using an abstracted model which incorporates dilated convolution and gated convolution. Then, a replication mechanism is incorporated into the generated model to ensure that both primary and secondary information are taken into consideration, while also optimizing the cluster search algorithm. Finally, the network structure is reconfigured in the generated model to effectively integrate the coding capabilities of the two-way language model and the text generation abilities of the one-way language model. We conducted experiments on the CNewSum dataset, and achieved Rouge-1, Rouge-2, and Rouge-L values of 44.21, 27.52, and 39.03 for the two-paragraph text summary model respectively. The results indicate a significant enhancement in the performance of the two-paragraph text summary model compared to the benchmark model.