Real-world learning is important because it encourages learners to obtain knowledge through various experiences. To design effective real-world learning, it is necessary to analyze the diverse learning activities that occur in real-world learning and to develop effective strategies for learning support. By inventing the technologies of multimodal sensing and knowledge externalization, we propose a method to capture the time-series occurrence of real-world learning and to analyze the spatial characteristics of a learning field that draws out diverse intellectual interactions. Our method generates multi-perspective indexes of the structure of a real-world learning field from the viewpoint of the occurrence probability of important learning behavior and knowledge acquisition. Our data analysis by this method extracted the regions where abundant discovery knowledge was not found and learners did not succeed in acquiring knowledge. Our method is useful for redesigning on-site and off-site activities in real-world learning, and for formatively improving the learning environment of real-world learning.