石家庄学院学报

为从海量数据中高效提取并整合产品文案元素，获得准确、个性化的搜索结果，提出基于知识图谱的产品文案元素搜索方法。该方法包括爬虫及索引模块、知识图谱模块和搜索模块。其中，爬虫及索引模块负责从互联网抓取相关数据，解析提取实体后存储并建立索引；知识图谱模块基于这些实体，通过知识抽取、融合对齐与加工，构建结构化的产品文案元素知识库；在搜索模块中，首先利用本体库解析用户查询关键词，然后将查询发送到知识图谱模块，随后利用Dijkstra算法在知识图谱中找到与用户查询最相关的产品文案元素的最短路径，从而优化搜索结果。实验结果表明：与上下文融合方法和双关系预测方法相比，所提出的方法搜索速度提升至1.8 s，准确率提升至95%，召回率提升至90%,F1分数高达0.92，用户体验获得了9.2的高评分，充分验证了方法的有效性和优越性。

In order to extract and integrate product copy elements from massive data efficiently and provide users with accurate and personalized search results, a search method of product copy elements based on knowledge graph has been proposed. This method design includes a crawler and indexing module, a knowledge graph module,and a search module. The crawler and index module grabs relevant data from the Internet, parses and extracts entities, stores them, and creates indexes. The knowledge graph module constructs a structured product copy element knowledge base based on these entities through knowledge extraction, fusion alignment, and processing. In the search module, the ontology library is first used to parse user query key words, and then the query is sent to the knowledge graph module. Subsequently, the Dijkstra algorithm is used to find the shortest path of product copy elements that are most relevant to the user's query in the knowledge graph to optimize the search results. The experimental results show that compared with the context fusion method and the dual relationship prediction method, the proposed method improves the search speed of product copy elements to 1.8s, the accuracy rate is increased to 95%,the recall rate is increased to 90%, the F1 score is up to 0.92, and the user experience score is up to 9.2, which fully verifies the effectiveness and superiority of the proposed method.