Temporal and Geographic Oriented Event Retrieval for Historical Analogy

There are many benefits to studying history. Recently, the study of support for learning history has emphasized the development of the ability to use knowledge of the past in an analogous manner when considering solutions to problems that arise in the present day, a process called historical analogy. Although previous studies have developed the ability of historical analogy using card games or datasets prepared in advance by experts, the situation is not ready for anyone to learn about a subject that is in line with their interests. In this study, we propose a Twitter chatbot that presents past events recorded on Wikipedia. When this chatbot receives a reply from a Twitter user, it analyzes the user's past tweets to collect geographic and temporal information of interest and returns past events that are close to those. We conducted experiments to confirm the effectiveness of our algorithm, and confirmed that the accuracy of our algorithm was approximately 30% points higher than that of the methods used in previous studies.