In an IoT solution using human activity recognition, devices that can be used may differ depending on the activity to be recognized. If we use the schema that has been proposed in previous studies for recommending combinations of solutions and devices, we cannot recommend the right devices for the activity, and the definition of solutions is duplicated. In this paper, we propose an ontology schema that connects the relationship between solutions and devices through a hierarchical concept of human activities. In addition, we propose a method for creating personalized recommendation candidates using it. As a result, we confirmed that the proposed method could represent the relationship between solutions and devices without duplication and recommend combinations of solutions and devices that fit the user's required activities.

Link prediction in knowledge graphs is used for question answering, for example, using the knowledge represented in the graph structure. On the other hand, there are still no established techniques for explaining inferences drawn from knowledge graphs. In this paper, we propose a technique for explaining the prediction results of link prediction using knowledge graph embedding. This explanation technique is realized by showing the characteristic parts of the predicted node's neighborhood.

Most of the daily living studies require cameras, sensor networks, and experimental space, and it is not easy to collect data by flexibly changing the conditions. In the future, for analysis of daily life, it is necessary to combine data obtained from both a physical environment that can acquire real data and a virtual environment that can flexibly change conditions and perform a large number of experiments. Moreover, in addition to recognizing activities from the collected data, it becomes possible to analyze daily living activities at the semantic level by adding abstract and background knowledge. In this study, we propose a method to construct knowledge graphs based on the simulation results of daily living using virtual space, in order to enable various analyzes of activities of daily living. First, we constructed an ontology representing activities and situations in virtual space, then constructed knowledge graphs of the daily living simulation results based on the ontology. Moreover, we proposed a method for knowledge graph augmentation by combining multiple activity graphs. This paper also discusses the knowledge graph generation method, proposed ontology, and potential for expansion.

In recent years, there has been increasing interest in explainable AI. When designing a white-box machine learning model, it is important not only to increase the prediction accuracy of the model, but also to design a model with high interpretability. In this paper, we consider the case in which a knowledge graph is used as a resource for learning data, and propose a method for selecting explanatory variables with high interpretability by utilizing the hyponymy relations between entities. We further confirmed with a test using a dataset of professional baseball players that the method can indeed choose the desired explanatory variables.

The use of knowledge graph technologies has been developed recently in historical studies, especially in the fields of prosopography or biography. The materials used in these studies, however, are mostly brought by secondary findings which have been given by the traditional research methods. This means that these studies do not necessarily deal with primary-source documents. Nevertheless, if we pursue more useful and profound application of knowledge graph technologies into historical studies, it should be necessary to consider the way to describe the very contents of primary-source documents. Although Factoid model already makes this kind of description partially possible, it is still not sufficient to express the chronological context and ambiguity which we find quite often in historical documents. Therefore, in this paper, we first propose the extension of the Factoid model with which we are now able to describe the ambiguous chronological context, and then introduce our data, constructed from ancient historical documents, and query examples.

Observing and analyzing human behavior is a crucial task in the domains of user-centered manufacturing. Video data is an important resource that captures the multimodal features of human behaviors. However, the analysis by manual is still a labor-intensive task such as extraction of interest scene. This paper proposes an ontology to represent human behavior in the video data as a knowledge graph to index the scene of interest.