A Novel Extractive Summarization Method Based on Multi-Path Feature Learning
Junzhi Huang · 2024
BERT (Devlin et al., 2018), a pre-trained Transformer (Vaswani et al., 2017) model, has achieved ground-breaking performance on multiple NLP tasks. In this paper, we introduce a text summarization approach based on multi-path feature learning, where the output of Bert is processed through three different pathways—RNN, GRU, and LSTM—before being utilized for summary extraction. For comparison, we use Transformer and BertSum as baselines. Experimental results demonstrate that our model achieves higher scores in terms of the ROUGE F1 evaluation metric on the CNN/DailyMail dataset.