With the rapid development of Semantic Web technologies, various knowledge graphs are published on the Web using Resource Description Framework (RDF), such as Wikidata and DBpedia. Knowledge graphs provide for setting RDF links among different entities, thereby forming a large heterogeneous graph, supporting semantic search, question answering and other intelligent services. Meanwhile, public availability of visual resource collections has attracted much attention for different Computer Vision (CV) research purposes, including visual question answering, image classification, object and relationship detection, etc. And we have witnessed promising results by encoding entity and relation information of textual knowledge graphs for CV tasks. Whereas most knowledge graph construction work in the Semantic Web and Natural Language Processing (NLP) communities still focus on organizing and discovering only textual knowledge in a structured representation. There is a relatively small amount of attention in utilizing visual resources for KG research. A visual database is normally a rich source of image or video data and provides sufficient visual information about entities in KGs. Obviously, making link prediction and entity alignment in wider scope can empower models to make better performance when considering textual and visual features together.
As mentioned above, general knowledge graphs focus on the textual facts. There is still no comprehensive multi-modal knowledge graph dataset prohibiting further exploring textual and visual facts on either side. To fill this gap, we provide a comprehensive multi-modal dataset (called Richpedia) in this paper, as shown in figure below.
In summary, our Richpedia data resource mainly makes the following contributions:
You can download images and triples of relationship from here through BaiduYun Drive. Because the image entity folder is relatively large, we split it into two parts(City&Sight, People) for download.
Our data uses other resources, so we make a statement here.
This work is licensed under a Creative Commons Attribution 4.0 International License
此处可能存在不合适展示的内容,页面不予展示。您可通过相关编辑功能自查并修改。
如您确认内容无涉及 不当用语 / 纯广告导流 / 暴力 / 低俗色情 / 侵权 / 盗版 / 虚假 / 无价值内容或违法国家有关法律法规的内容,可点击提交进行申诉,我们将尽快为您处理。