通过重组事件片段提升长文档摘要质量，无需额外训练。
HERA: Improving Long Document Summarization using Large Language Models with Context Packaging and Reordering
- 按语义分段并聚合同一事件内容，再重新排序形成清晰上下文。
- 在两个数据集上超越基础模型，ROUGE和忠实度指标均提升。
- 无需微调或额外资源，适合快速部署到现有系统中。
尽管大型语言模型（LLMs）的上下文长度迅速增长，其在长文档摘要任务中的表现仍不理想。一个重要原因是事件相关信息在长文档中分散分布，且叙述顺序杂乱，影响了LLMs对长文档的准确理解与利用。为此，我们提出一种新型摘要生成框架HERA：首先根据语义结构对长文档进行分段，检索属于同一事件的文本片段，并最终重新排序以构建输入上下文。我们在两个长文档摘要数据集上评估该方法，实验结果表明，HERA在ROUGE、BERTScore和忠实度指标上均优于基础模型，且无需额外微调或资源投入。
原文摘要 · Abstract (English)
Despite the rapid growth of context length of large language models (LLMs) , LLMs still perform poorly in long document summarization. An important reason for this is that relevant information about an event is scattered throughout long documents, and the messy narrative order impairs the accurate understanding and utilization of LLMs for long documents. To address these issues, we propose a novel summary generation framework, called HERA. Specifically, we first segment a long document by its semantic structure and retrieve text segments about the same event, and finally reorder them to form the input context. We evaluate our approach on two long document summarization datasets. The experimental results show that HERA outperforms foundation models in ROUGE, BERTScore and faithfulness metrics, while HERA does not require additional fine-tuning and resources.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。