نوع مقاله : علمی پژوهشی
نویسندگان
1 استاد گروه علم اطلاعات و دانش شناسی، دانشکده علوم تربیتی و روانشناسی، دانشگاه شهید چمران اهواز، اهواز، ایران
2 دکتری، گروه علم اطلاعات و دانش شناسی، دانشکده علوم تربیتی و روانشناسی، دانشگاه شهید چمران اهواز، اهواز، ایران
3 دکتری علم اطلاعات و دانش شناسی، دانشگاه قم، ایران
کلیدواژه‌ها
عنوان مقاله English
نویسندگان English
Background and Objectives: Digital libraries generate large volumes of transaction and search-log data that can reveal users’ actual information needs and support evidence-based collection development, digitization priorities, resource organization, and personalized services. The Ayatollah Boroujerdi Digital Library in Qom holds distinctive collections, including manuscripts, hot-metal and lithographic publications, textbooks, theses, and other rare resources. Despite the value of these collections, conventional usage statistics provide limited insight into relationships among the subjects searched by users. This study therefore aimed to identify clusters of subject use and association patterns among users’ searches in the library’s digital databases. More specifically, it examined which subjects were frequently used together and how the resulting patterns could contribute to acquisition, collection development, further digitization, improved service quality, and the future design of recommender systems.
Methodology: This cross-sectional analytical study examined ten years of digital-library usage data covering the period 2011–2021. The dataset was extracted from the NOSA digital-resource archive system. The original databases were stored in SQL format and the extracted records were converted to CSV files for analysis. Data preparation and cleaning included separating date and time values recorded in a single field, converting ordered subject labels into codes, and removing records associated with IP addresses generating more than 50 searches. This procedure was intended to distinguish searches by individual users from activity originating from shared systems or internal library computers. Subjects assigned to each resource were retrieved through resource codes and consolidated into principal subject categories. A two-step clustering procedure was applied because it is suitable for large datasets and can accommodate both categorical and quantitative data. Association-rule mining using the Apriori algorithm was then employed to identify dependencies and co-occurrence patterns among searched subjects. IBM Modeler version 18 was used for clustering and association-rule analysis. For network visualization, the data were transformed into a matrix and analyzed with the bibliometrix package in R. The clustering analysis concerned searched subjects rather than the physical type or digital format of resources.
Findings: PDF was the dominant format, accounting for 98.35% of the resources used. Manuscripts represented 99.91% of the records for which resource type was identifiable, demonstrating the central role of this collection in digital-library use. According to the recorded counts, 709,337 of the 1,048,575 visits, approximately 68%, were made by guest users, whereas 339,238 were associated with non-guest users. Overall use increased markedly over the study period, reaching 245,187 recorded uses in 2018 and 244,333 in 2020. December, April, and January had the highest aggregated use, and the busiest daily period was between 10:00 a.m. and 12:00 noon. Two subject clusters were identified. Cluster 1 covered approximately 56% of the data and was characterized by jurisprudence, principles of jurisprudence, prayer, literature, medicine, mysticism, history, ethics, and sciences. Cluster 2 accounted for approximately 44% and included theology, Hadith, syntax, philosophy, logic, poetry, astronomy, prayers, interpretation, and miscellaneous subjects. Network analysis showed that jurisprudence, syntax, theology, principles of jurisprudence, and Hadith were among the most prominent and highly connected subjects. Jurisprudence had particularly strong co-search relationships with logic, ethics, medicine, Hadith, history, and syntax. The association rules showed confidence values of approximately 55% for some relationships involving jurisprudence, most notably its relationships with logic, ethics, and medicine. By contrast, subjects such as divination, elegies concerning martyrdom, commentary, and arithmetic attracted comparatively little attention.
Discussion: The findings demonstrate that digital-library logs can provide managers with evidence of actual user behavior that is not readily available through surveys or aggregate visit statistics. The predominance of manuscripts and the concentration of use in the first subject cluster suggest that digitization, description, and collection-development activities should give particular attention to jurisprudence and its related fields. At the same time, the relatively limited use of the library’s rare hot-metal and lithographic holdings indicates a need for better online promotion, standardized organization, and improved discoverability. Because a large proportion of users access the library as guests, encouraging optional membership could make it possible to analyze longitudinal interests and provide more personalized services without restricting public access. The co-search patterns can also inform subject navigation, “related resources” features, and recommender systems. Nevertheless, the results should be interpreted as associations rather than causal relationships. They reflect one library, depend on the quality of subject coding and log data, and do not incorporate resource format into the clustering model. Future research should compare Apriori with techniques such as k-means and test combined clustering and association-rule approaches in other digital-library environments.
کلیدواژه‌ها English
