📄 中文摘要
大型语言模型（LLMs）在科学创意生成领域展现出潜力，但生成的结果往往缺乏可控的学术背景和可追溯的灵感路径。为了解决这一问题，提出了一种名为GYWI的科学创意生成系统，该系统结合了作者知识图谱与检索增强生成（RAG），形成外部知识库，为LLMs提供可控的背景和灵感路径追踪。首先，提出了一种以作者为中心的知识图谱构建方法和灵感源采样算法，以构建外部知识库。然后，提出了一种混合检索机制，由RAG和GraphRA组成，以增强生成过程的有效性。
通过图谱激发灵感：将合著者图谱与检索增强生成相结合，以促进基于大型语言模型的科学创意生成
📄 English Summary
Graph Your Way to Inspiration: Integrating Co-Author Graphs with Retrieval-Augmented Generation for Large Language Model Based Scientific Idea Generation
Large Language Models (LLMs) show promise in the domain of scientific idea generation; however, the outputs often lack controllable academic context and traceable pathways of inspiration. To address this issue, a scientific idea generation system named GYWI is proposed, which integrates author knowledge graphs with retrieval-augmented generation (RAG) to create an external knowledge base that provides controllable context and traces of inspiration paths for LLMs to generate novel scientific ideas. An author-centered knowledge graph construction method and inspiration source sampling algorithms are introduced to build the external knowledge base. Additionally, a hybrid retrieval mechanism is proposed, consisting of both RAG and GraphRA, to enhance the effectiveness of the generation process.