We conduct fundamental research on machines for learning systems from data. Natural language and knowledge are well-known examples of the most challenging targets to handle on machines. One of the ultimate goals of our research is to elucidate the essence and to establish a methodology in which machines acquire and utilize natural language and knowledge as effectively and efficiently as human beings.

While artificial intelligence (AI)-related technology has become popular, we encounter the black-box problem; the problem is that humans are difficult to explicitly interpret how deep learning models acquire and store necessary information from data and how they leverage learned clues effectively during the computation. Similarly, new research issues have arisen due to the development of AI technologies, such as issues related to fairness resulting from the bias of data, and issues related to fake information generated by misuse of AI technology. We aim to verify and analyze various old and new issues in AI-related technologies theoretically and empirically and reveal the principle and essence of those issues.