A Topical Keywords Fusion Based on Transformer For Text Summarization
Shuai Zhao, Fucheng You · 2020
In recent years, the model of sequence to sequence based on attention mechanism have made impressive progress in abstractive document summarization. Unfortunately, as abstractive summarization research is in a primitive stage, the performance of these models is still far from standard. In this paper, we propose a novel method called Abstractive Summarization with Transformer and Topical Keywords(ASTTK). This method augments the attention sequence-to-sequence model in two aspects. First, we use the Transformer as feature extractor to improve the performance of the model, which alleviates the poor parallel computing ability or poor remote feature capture ability of the traditional model. Second, we creatively integrate the topical keyword information, which too makes the generated summary more close to the standard summary. Experiments on data show that the model has a significant improvement in ROUGE index and readability, which shows that the proposed model can integrate keyword information effectively and generate core abstracts.