Design on Weak Supervision and Self-learning Spatiotemporal Data Annotation System for Large-scale Complex Scenes

All kinds of news reports cover a large number of important social events, which are presented to readers in the form of images and texts. By tagging the time, locations, targets and events of texts and images, rich tagging data are generated, which can provide the data basis for analyzing and mining spatiotemporal events. The designed "weak supervision and self-learning spatial data annotation system for large-scale complex scenes" provides two ways for texts and images: weak supervision manual annotation and self-learning automatic batch annotation. The system does not only realize the automatic annotation of batch data, but also supports for manually tagging the data with recommendation. At the same time, it can be applied to teaching and learning, so that students can better understand the principle and process of annotation.