基于多模态信息相关度计算的计算机教育领域实体链接

Entity Linking in the Field of Computer Education Based on Multi-modal Information Relevance Calculation

教育领域通常将学科中的知识体系联系起来构建学科知识图谱,但由于其知识体系并非一成不变,因而学科知识图谱的更新显得尤为重要,而实体链接作为构建和更新知识图谱的一项基础技术尤为关键。为了实现计算机学科领域多模态知识图谱的更新,提出一种视觉实体链接方法。首先利用YOLOV5识别图像中的视觉实体,结合视觉实体提取并筛选图像文本描述中的三元组;然后面向计算机学科领域提出一种视觉实体链接规则(VELR),为了有效完成链接,依次利用图像视觉实体和筛选后的三元组中头尾实体以寻找最佳链接点,并为不同的链接点设置不同的链接策略,且规则中充分利用图像视觉实体和筛选后的三元组对链接后的实体进行扩展。三元组分类实验表明,提出的方法在计算机学科领域多模态知识图谱实现链接后产生的三元组是正确的,证明了所提方法的有效性。

In the field of education,the knowledge system in the discipline is usually linked to construct the discipline knowledge graph.How⁃ever,because the knowledge system is not immutable,it is particularly important to update the discipline knowledge graph.As a basic technol⁃ogy of building and updating knowledge graph,entity link is very critical.Therefore,in order to update the multi-modal knowledge graph in computer field,the paper proposed a visual entity linkage method.Firstly,we used YOLOV5 to identify the visual entities in the image.By combining visual entities,the triples in image text description were extracted and filtered.Secondly,we proposed a visual entity link rule(VELR)for the computer field.In order to effectively complete the link,the image visual entity and the head and tail entity of the filtered trip⁃let were used successively to find the best link point.We set different link strategies for different link points.The rules made full use of image visual entities and filtered triples to expand the linked entities.Finally,the triplet classification experiment shows that the proposed method is correct to realize the triplet generated by the linkage of multi-modal knowledge graph in the field of computer science,which proves the validi⁃ty of the proposed method.