The existing automatic text summarization methods have the problems of redundant key information and low accura-cy in processing long texts.This paper proposes a three-stage text summarization algorithm for Chinese long text.Firstly,the text compression algorithm is used to compress the long text information in a fixed range,and the redundant information irrelevant to the topic is filtered.Then,the deep semantic features of sentences are learned by combining the pre-training model Bert,and the key sentences with rich topic information are further extracted.Finally,the Seq2Seq model with the pointer mechanism is used to rewrite the key sentences.The effectiveness of this algorithm on Chinese long text summarization is verified by experiments based on real large-scale long text data in financial field.
关键词
摘要生成/长文本压缩/预训练/摘要生成/指针机制
Key words
text summarization/long text compression/pre-training/summary generation/pointer mechanism
维普
万方数据