基于双向GRU与Attention机制的改进Seq2Seq自动文本摘要模型研究

A Bidirectional GRU and Attention-Based Improved Seq2Seq Model for Automatic Text Summarization

摘要: 随着大数据时代的到来，数据规模呈指数级增长，信息过载问题日益突出。自动文本摘要技术通过计算机模型提炼文本主旨，生成简洁摘要，能够有效缓解信息过载，并广泛应用于新闻标题生成、文本检索与智能问答等场景。本文在分析现有文本摘要技术的基础上，以Seq2Seq模型为研究核心，重点探讨注意力机制(Attention Mechanism)在摘要生成中的作用，并提出一种结合双向GRU与注意力机制的改进型Seq2Seq模型。该模型在编码器部分采用双向GRU (BiGRU)结构，以充分捕获上下文语义信息；在解码器部分引入注意力机制，以提升摘要生成的准确性与连贯性。本文基于CNN/Daily Mail英文数据集对所提出模型进行训练与测试，并采用ROUGE指标评估实验结果。实验表明，所提出模型在摘要质量与信息覆盖度方面均优于传统Seq2Seq模型，验证了其有效性与可行性。

Abstract: With the advent of the big data era, data volumes are growing exponentially, and the problem of information overload has become increasingly prominent. Automatic text summarization technology, which distills the main ideas of a text through computational models to generate concise summaries, can effectively alleviate information overload and is widely applied in scenarios such as news headline generation, text retrieval, and intelligent question answering. Based on an analysis of existing text summarization techniques, this paper focuses on the Seq2Seq model and explores the role of the Attention Mechanism in summary generation. A novel improved Seq2Seq model combining bidirectional GRU (BiGRU) and attention mechanism is proposed. In this model, the encoder employs a BiGRU structure to fully capture contextual semantic information, while the decoder incorporates an attention mechanism to enhance the accuracy and coherence of generated summaries. The model is trained and tested on the CNN/Daily Mail English dataset, and evaluation is conducted using the ROUGE metric. Experimental results demonstrate that the proposed model outperforms the traditional Seq2Seq model in both summary quality and information coverage, validating its effectiveness and feasibility.