School libraries can support teachers by selecting a set of books for classes, called teaching material curation. While increasing demand for inquiry-based learning attaches more importance to it, existing library tools have not supported it well. We developed a system to support teaching material curation, BookReach, which lets users obtain candidate books for teaching material by selecting the textbook chapter of a requested class. BookReach displays the covers of candidate books from the user's school library collection based on a table that maps relevant decimal classification classes to textbook chapters. We evaluated the system with three usability perspectives: effectiveness, efficiency, and satisfaction. We asked 18 school librarians to curate books for mock classes by using BookReach. The results showed the relevance of candidate books, less cost in time than OPAC, and user-friendliness in operation. The librarians' free-text questionnaire feedback revealed requests for book-display configuration and collection data editing.

　The purpose of this study was 1) to understand the actual situation of reference relationships among the rules in the 2018 edition of the Nippon Cataloging Rules (NCR2018) being expressed as RDF (Resource Description Framework) data, by aggregating the reference relations, and 2) to grasp the characteristics at a global level by applying network analysis to those relations. 1) To identify the rules that should be adopted as reference destinations from the reference instructions given in the form of a range specification, we devised our own methods. 2) The reference relationships extracted from Chapters 2, 4, and 6 of NCR2018 were aggregated in terms of rule numbers, such as the percentage of rules that have references. 3) The number of nodes, edges and their occurrences were counted for each graph that was formed from the extracted reference relationships, and various features of the graphs were calculated to summarize the features of each graph. 4) The centrality of the graphs was determined; using four measures, the top nodes were extracted and integrated, and then, hierarchical clustering was performed to confirm the contents of the resultant clusters. These results gave us the actual state of NCR2018 reference relationships with specific numeric values, and the big-picture characteristics of those relationships.