基于BERT模型的创新知识图谱构建研究

Construction of an Innovation Knowledge Graph Based on the BERT Model

摘要: 本文提出了一种基于BERT模型的创新知识图谱构建方法，旨在从大规模文本数据中自动抽取高质量的实体及其关系。首先，利用BERT强大的上下文语义表示能力，提取论文、专利中的关键信息，通过核心实体识别、关系抽取与链接，构建创新知识图谱，并开展创新指数分析。通过构建创新知识图谱，实现对技术创新信息的结构化表达与语义关联，方便后续的知识推理与应用。结果显示，基于Bert模型的创新知识图谱能够有效地提取创新指数，并分析产业上下游联系。

Abstract: This paper proposes a method for constructing an innovation knowledge graph based on the BERT model, aiming to automatically extract high-quality entities and their relationships from large-scale textual data. Firstly, leveraging BERT’s powerful contextual semantic representation capabilities, key information is extracted from papers and patents through core entity recognition, relation extraction, and linking to build the innovation knowledge graph, followed by innovation index analysis. By constructing the innovation knowledge graph, the method achieves structured representation and semantic correlation of technological innovation information, facilitating subsequent knowledge reasoning and applications. The results demonstrate that the BERT-based innovation knowledge graph can effectively extract innovation indices and analyze upstream and downstream industrial connections.